You are on page 1of 8

Estimating storm erosion with a rainfall simulator

S. C. Nolan1, L. J. P. van Vliet2, T. W. Goddard1, and T. K. Flesch3


1Conservation and Development Branch, Alberta Agriculture, Food and Rural Development, #206, 7000-113
Street, Edmonton, Alberta, Canada T6H 5T6, e-mail: snolan@ulysses.sis.ualberta.ca; 2British Columbia Land
Resource Unit, Agriculture and Agri-Food Canada, Agassiz, British Columbia, Canada V0M 1A0; 3Department of
Earth and Atmospheric Sciences, University of Alberta, Edmonton, Alberta, Canada T6G 2H4. Received
27 August 1996, accepted 10 July 1997.

Nolan, S. C., van Vliet, L. J. P., Goddard, T. W. and Flesch, T. K. 1997. Estimating storm erosion with a rainfall simulator.
Can. J. Soil Sci. 77: 669–676. Interpreting soil loss from rainfall simulators is complicated by the uncertain relationship between
Can. J. Soil. Sci. Downloaded from cdnsciencepub.com by 130.105.110.3 on 09/04/22

simulated and natural rainstorms. Our objective was to develop and test a method for estimating soil loss from natural rainfall using
a portable rainfall simulator (1 m2 plot size). Soil loss from 12 rainstorms was measured on 144-m2 plots with barley residue in
conventional tillage (CT), reduced tillage (RT) and zero tillage (ZT) conditions. A corresponding “simulated” soil loss was cal-
culated by matching the simulator erosivity to each storm’s erosivity. High (140 mm h–1) and low (60 mm h–1) simulation inten-
sities were examined. The best agreement between simulated and natural soil loss occurred using the low intensity, after making
three adjustments. The first was to compensate for the 38% lower kinetic energy of the simulator compared with natural rain. The
second was for the smaller slope length of the simulator plot. The third was to begin calculating simulator erosivity only after
runoff began. After these adjustments, the simulated soil loss over all storms was 99% of the natural soil loss for CT, 112% for
RT and 95% for ZT. Our results show that rainfall simulators can successfully estimate soil loss from natural rainfall events.

Key words: Natural rainfall events, simulated rainfall, erosivity, tillage

Nolan, S. C., van Vliet, L. J. P., Goddard, T. W. et Flesch, T. K. 1997. Estimation de l’érosion pluviale au moyen d’un simu-
For personal use only.

lateur de pluie. Can. J. Soil Sci. 77: 669–676. L’interprétation des déperditions de sol à partir de données obtenues au moyen de
simulateurs de pluie est rendue compliquée par l’incertitude des rapports existants entre pluie simulée et pluie naturelle. L’objet
de nos travaux était de mettre au point et de tester une méthode d’estimation des déperditions de sol causées par les pluies
naturelles au moyen d’un simulateur de pluie portatif (sur parcelle de 1 m2). Les déperditions de sol résultant de 12 épisodes de
pluie étaient mesurées dans des parcelles de 144 m2 portant des restes de culture d’orge et conduites en régime de travail du sol
classique (TC), de travail réduit (TR) et de cultures sans travail du sol (CST). Les déperditions correspondantes en conditions
simulées étaient calculées en appariant les valeurs d’érosivité du simulateur au pouvoir érosif réel de chaque épisode pluvial. On
utilisait une forte (140 mm h–1) et une basse (60 mm h–1) intensité de précipitations simulées. La meilleure concordance entre les
déperditions de sol causées entre les pluies réelles et les pluies simulées s’observait au régime d’intensité pluviale inférieure,
moyennant toutefois 3 corrections, la première, pour compenser la moindre (38 %) énergie cinétique du simulateur par rapport aux
précipitations naturelles; le second pour tenir compte de la pente plus courte de la parcelle sous simulateur de pluie, tandis que la
troisième consistait à ne commencer à calculer l’érosivité en régime simulé qu’après le début du ruissellement. Moyennant ces
trois corrections, les déperditions de sol en simulation, tous épisodes pluviaux confondus, correspondaient à 99 % des déperditions
par pluie naturelle observées en régime TC, à 112 % en régime TR et à 95 % en régime CST. Il semble que les simulateurs de
pluie permettent parfaitement d’estimer les déperditions de sol causées par les épisodes de pluie naturelle.

Mots clés: Épisodes de pluie naturelle, pluie simulée, érosivité, travail du sol

Rainfall simulators play an important role in understanding at this question but focused on relative measures of erosion,
soil erosion. They allow soil loss and runoff to be generated such as soil loss per unit of runoff, not the amount of soil
under controlled and repeatable conditions. However, the loss. Two difficult problems exist in relating soil loss from
interpretation of simulator measurements is complicated by simulated and natural rainfall: replicating the erosive energy
the uncertain relationship between the erosiveness of simu- (erosivity) of a specific storm with a simulator, and “scal-
lated and natural rainstorms. A common operating proce- ing-up” soil loss from the typically small simulator plots to
dure is to choose a precipitation intensity, then run the larger field-scales.
simulator for an essentially arbitrary time (Andraski et al. One approach to replicating the erosivity of natural
1985), or until steady state runoff is achieved (Elliot et al. storms is to duplicate the intensity and duration of a natural
1989; West et al. 1991). Although the resulting measure- storm with a simulator. Besides being impractical, this
ments are informative, they could be given greater context if ignores important differences between natural and simulat-
related to historic storms. ed rainfall properties that alter erosivity, such as drop size
Can soil loss generated with a rainfall simulator approxi- distribution and drop velocities (Peterson and Bubenzer
mate field-scale soil loss from specific storms? McIsaac and 1986). A more practical approach is to use extreme rainfall
Mitchell (1992) and Robichaud and Waldrop (1994) looked climatology to select a simulation duration and intensity that
669
670 CANADIAN JOURNAL OF SOIL SCIENCE

can be statistically related to a historic event, such as a one


in 100 yr storm (Blough et al. 1990). Yet how well do such
statistical “duration-intensity” storms translate into historic
storm erosivities? The problem of differences in natural and
simulated rainfall characteristics would also remain.
Even if the erosivity of natural rainfall can be approxi-
mated with a rainfall simulator, one must “scale-up” simu-
lated soil loss to estimate field-scale soil loss. With
small-plot simulators there are no realistic means of gener-
ating the rill erosion found on long slopes. Foster et al.
(1981b) indicate that only sheet erosion can be measured
from plots of less than about 4 m in length. Soil loss from a
simulator will then underestimate erosion from larger areas.
Other difficulties when comparing soil loss from the simu-
Can. J. Soil. Sci. Downloaded from cdnsciencepub.com by 130.105.110.3 on 09/04/22

lator with natural erosion include the question of how many


small plot simulator replicates are needed to characterize
field scale soil loss, and how to account for differences
between antecedent conditions of natural and simulated ero-
sion events.
The objective of our study was to develop and test a Fig. 1. The GRS-II rainfall simulator used in this study.
method for estimating soil loss from natural rainstorms
using measurements from a portable rainfall simulator. Our
procedure was based on matching the erosivities of simulat- Soil Loss Measurements
ed and natural rainfall, using Wischmeier and Smith’s Soil loss from natural rainstorms (referred to as “natural”
(1978) R factor. Adjustments for differences in the kinetic soil loss) was collected from the 144 m2 plots between May
energy of simulated and natural rainfall, and for differences 1987 and June 1991. Measurements were made at the end of
For personal use only.

in the slope lengths of the simulated and natural plots were each spring season (defined as after snowmelt, but before
also used. We compared soil loss from a rainfall simulator crop establishment). These five measurement periods
on 1 m2 plots with natural soil loss from 12 rainstorms mea- included a total of 12 rainstorms (Table 1). Precipitation
sured on 144 m2 plots. Soil loss was measured under CT, RT
intensity and duration were recorded with a Belfort
and ZT conditions previously cropped to barley.
Universal recording rain gauge. Snow accounted for 6% of
MATERIALS AND METHODS the precipitation (water equivalent) during the 1989 period,
and 9% during 1990. Soil loss from this snowmelt was con-
Study Site sidered slight, and the snowfall was ignored in the erosivity
Our study site was a long-term barley tillage experiment calculations.
near Dawson Creek, British Columbia. The site was a 6–8% Soil loss from the rainfall simulator (referred to as “sim-
sloping Donnelly silt loam (Solonetzic Gray Luvisol). ulated” soil loss) was measured on 27–30 May 1991, before
Erosion was measured from bounded 30.0 m (upslope) by spring tillage. Table 2 outlines the soil conditions at the time
4.8 m plots in CT, RT and ZT conditions. Each tillage con- of simulation. Sites were not pre-wetted, and no precipita-
dition was replicated three times in a randomized complete tion occurred during the simulations. Low- and high-inten-
block design. Experimental and site details are described in sity simulations were conducted on separate 1 m2 plots,
van Vliet et al. (1993). within each of the 144 m2 natural soil loss plots. Each sim-
ulation lasted approximately 20 min. Runoff subsamples
Rainfall Simulator
were collected by suction every 3 min after runoff started.
A tripod mounted Guelph Rainfall Simulator (GRS II)
(Tossell et al. 1987) was used to generate soil loss. The All remaining runoff was collected by suction and by wash-
portable GRS II delivers water from a single nozzle sus- ing the collection apron after each simulation. The subsam-
pended above the ground (Fig. 1). Soil loss was collected ple measurements were used to proportion the total soil loss
from bounded 1.0 m by 1.0 m plots within a wetted perime- over the simulation. Cumulative soil loss “curves” were cre-
ter that extended approximately 0.5 m beyond the plot bor- ated by linear interpolation between the subsample mea-
ders. We used a low-intensity simulation targeted at 60 mm surements.
h–1, using a 3/8 inch 20W nozzle, at a height of 1.2 m and
pressure of 69 kPa (Table 2 in Tossell et al. 1987). A high- Estimating Natural Soil Loss from Simulated
intensity simulation was targeted at 140 mm h–1, using a 1/2 Rainfall
inch 30W nozzle, at a height of 1.7 m and pressure of 48 kPa MATCHING EROSIVITY. Our approach was to replicate natural
(Tossell et al. 1987). The low-intensity simulations varied rainstorm erosivity with the rainfall simulator. Our measure
by ±10% between simulations and the high-intensity ones of erosivity was the R (erosivity) factor of Wischmeier and
by ±19%. Further details of the simulator study are present- Smith (1958). The R factor of a natural storm is determined
ed in Nolan et al. (1992). by the precipitation characteristics, and is calculated as:
NOLAN ET AL. — STORM EROSION WITH A RAINFALL SIMULATOR 671

Table 1. Natural rainstorm characteristics and erosivity (R) over the study period
Storm Storm Max. 30-min Measurement
Measurement Storm precipitation duration intensity Storm R period R
period date (mm) (h) (mm h–1) (MJ mm ha–1 h–1) (MJ mm ha–1 h–1)
1 20 June 1987 26.2 13.0 12.8 56.5 56.5

2 13 May 1988 19.0 10.0 7.0 22.5


5 June 1988 73.5 30.0 15.0 199.4 221.9

3 20 May 1989 22.0z 24.0 3.0 8.7


8 June 1989 17.0 12.0 10.0 26.9
9 June 1989 9.0 0.5 18.0 32.8
13 June 1989 27.0 4.0 34.0 193.4 261.8

4 17 April 1990 12.5 9.0 8.0 14.5


24 May 1990 24.0 15.0 8.0 28.8
Can. J. Soil. Sci. Downloaded from cdnsciencepub.com by 130.105.110.3 on 09/04/22

10 June 1990 104.0 50.0 6.0 98.0 141.3

5 13 June 1991 14.0 10.0 8.0 7.9


4 June 1991 31.0 20.0 12.0 27.5 35.4
z 1.7% of precipitation as snow.

Table 2. Soil surface conditions at the time of the rainfall to twice the amount of rain (although not dimensionally cor-
simulations rect), thus:

Rsim = Esim 2 Psim = Esim 2( Dsim Isim )


Gravimetric
Tillage Crop residue Roughness soil moisture Bulk density (4)
type cover (%) indexz (% by weight) (Mg m–3)
For personal use only.

Conventional 48b 126b 11.7a 1.1a where Esim is the “storm” energy of the simulated rainfall
Reduced 53b 335a 11.8a 1.1a (Eq. 2), Psim is the depth of water applied in the simulation
Zero 89a 195ab 9.0a 1.1a (mm), Dsim is the simulation duration (h), and Isim is the sim-
Standard Error 2.5 46 2.5 0.1
zRillmeter
ulated intensity (mm h–1). A duration Dsim could then be
method, index after Romkens and Wang (1986). chosen to give Rsim matching any of the 12 natural storms:
a,bValues within a column followed by the same letter are not signifi-
cantly different at P < 0.05 using Fisher’s protected LSD.
R
Dsim (5)
2 Isim Esim
R = EI30 (1)

where I30 is the maximum 30 min rainfall intensity of the The simulated soil loss at time Dsim, which will correspond
storm (mm h–1), and E is the total storm energy (MJ ha–1). to the natural soil loss for the given R, was determined from
Storm energy is determined by dividing the storm into n the cumulative soil loss curves created from the simulator
segments of approximately constant intensity: measurements. Therefore only one simulator run was used
to calculate simulated soil loss for each of the 12 storms. For
n
the low intensity simulations, the Dsim corresponding to 2 of
E= ∑ ei Pi (2) the 12 storms exceeded the 20 min simulation durations, and
i =1 we linearly extrapolated the cumulative soil loss curves (but
by no more than 1.8 min past our field measurements).
where ei is the energy per unit rainfall (MJ ha–1 mm–1) dur-
ing the segment, and Pi is the segment precipitation (mm). ENERGY ADJUSTMENT. An important assumption when calcu-
Foster et al. (1981a) give an empirical expression for ei: lating Dsim (Eq. 5) is equality between the kinetic energy of
simulated and natural rain at equal intensities. However,
ei = (0.119 + 0.0873 log10 Ii ) (3) Tossell et al. (1990) found that the kinetic energy of rainfall
from the GRS II was lower than natural rainfall, due to dif-
where Ii is the segment rainfall intensity (mm h–1); whenev- ferences in drop sizes and velocities. We calculated an ener-
er ei exceeds 0.283 MJ ha–1 mm –1, it is set at that value. gy adjusted Dsim by assuming that Esim was 38% of natural
Separate R values were calculated for each of the 12 natural rainfall at the low intensity and 45% of natural for the high
rainstorms in our study (Table 1). intensity (interpolating from Tossell et al. (1990)). For three
The 12 natural rainstorm R values were mimicked with of the 12 storms the adjusted Dsim values were larger than
the rainfall simulator (Rsim), using an adaptation of Eq. 1. the 20 min simulation durations for the low intensity, and
For rain durations less than 30 min, such as our simulations, the soil loss curves were linearly extrapolated by up to 32
Wischmeier and Smith (1978) indicate that I30 should be set min.
672 CANADIAN JOURNAL OF SOIL SCIENCE

SCALE ADJUSTMENT. When comparing erosion from plots five measurement periods, the natural soil loss was almost
with different slope lengths, it is important to consider the 10 times the simulated value. There was also a large “scat-
relative importance of sheet and rill erosion. Rill erosion ter” in the relationship between the two (r2 = 0.22).
increases with slope length (Foster et al. 1981b). As our Adjustments to account for the lower kinetic energy of
30 m long natural plots will produce more rill erosion than the simulator (E adjustment) and the smaller slope length (L
our 1 m long simulator plots, we adjusted the simulated soil adjustment) led to better agreement with natural soil loss for
loss upward to account for the “missing” rill erosion. the low-intensity simulations (in both correlation coefficient
The Universal Soil Loss Equation (USLE) (Wischmeier and magnitude). Figure 4 shows the result of these adjust-
and Smith 1978) provides a simple means of accounting for ments on the accumulated soil loss from all five measure-
the effect of slope length on erosion. In the USLE, soil loss ment periods. The unadjusted soil loss represented only 14
over the length of a uniform slope is proportional to a slope to 18% of the natural soil loss from the three tillage treat-
length factor, calculated as: ments. With the E and L adjustments, the low-intensity sim-
ulated soil loss rose to 64% of natural in CT (significantly
α
L = (λ / 22.1) (6) different), 76% in RT (not significantly different), and 47%
Can. J. Soil. Sci. Downloaded from cdnsciencepub.com by 130.105.110.3 on 09/04/22

in ZT (significantly different).
where λ is distance up-slope (m) from the origin of overland Using the high-intensity simulations gave an unadjusted
flow, 22.1 m is the USLE standard plot length, and the expo- soil loss that was not significantly different from natural soil
nent α increases as a soil is more susceptible to rill erosion. loss for CT and RT (Fig. 4). The unadjusted soil loss was
Foster et al. (1977) relate α to the ratio of rill to sheet ero- 117% of natural for CT, 87% for RT, although only 7% for
sion (β): ZT. The E and L adjustments worsened the agreement to as
much as a 623% overestimation of the natural soil loss.
α = β / (1 + β ) (7) We made an additional adjustment to determine our sim-
ulated soil loss. We commenced our calculations of simula-
where β is computed from McCool et al. (1989): tor erosivity when runoff began, rather than when the
simulation began (adjustment B). The desired duration Dsim
(
β = (sin θ / 0.0896) / 3(sin θ )0.8 + 0.56 ) (8) is then the simulator duration after runoff begins. This was
For personal use only.

based on comments by Wischmeier and Mannering (1969)


and θ is the slope angle. This expression for β is for condi- that runoff was almost instantaneous under natural rain-
tions where rill and sheet erosion are equal over a 22.1 m storms, but in simulated rainfall studies a substantial time
plot length (a typical agricultural soil). elapsed before runoff began. They disregarded this time lag
Our adjustment for the “missing” rill erosion was to mul- in their determination of the USLE soil erodibility (K) fac-
tiply the simulated soil loss by the ratio of L for the natural tor from simulated rainfall. Moreover, Auerswald et al.
plots to L of the simulator plots. Following the advice of (1992) found that the runoff lag increased with decreasing
McCool et al. (1989), we used λ = 4 m when calculating L simulator plot size, and recommended disregarding rainfall
for our simulator plots and λ = 30 m for the natural plots. over this time lag when calculating simulator erosivity. This
With an average slope of 7%, L for the simulated measure- adjustment required further extrapolation of soil loss from
ments is 0.46 and L for the natural plots is 1.15. We there- our low-intensity simulations for 4 of the 12 storms (up to
fore multiplied the simulated soil loss by 2.5 to adjust it to 21.6 min beyond our measurements).
the natural rainfall plot length. When adjusted for E, L and B, there was excellent agree-
ment between the natural and simulated soil loss using the
RESULTS low intensity simulation (Fig. 4). The E-L-B adjusted simu-
“Natural” soil loss was consistent with our expectations that lated soil loss was 99% of the natural soil loss for CT, 112%
CT > RT > ZT (Table 3). This ordering was generally seen for RT and 95% for ZT. None of these is statistically differ-
in the “simulated” soil loss (Fig. 2), although this depended ent from natural soil loss. However, the E-L-B adjustments
on the simulation time considered. Although we expected to the high-intensity simulations worsened the agreement to
CT would have the earliest runoff in our simulations, RT as much as 948% overestimation of natural soil loss.
was the earliest, followed by CT and ZT. In both the simu-
lated and natural cases, ZT was very effective in reducing DISCUSSION
soil loss. The accumulated soil loss estimated from the low-intensity
An initial comparison of natural and simulated soil loss simulator measurements was within 12% of the natural soil
for the five measurement periods (which included 12 loss for each tillage treatment for the five measurement peri-
storms) is shown in Fig. 3. The simulated soil loss was ods. This good agreement required a series of three adjust-
determined by adding the soil loss for each storm in the peri- ments in our calculations of simulated soil loss. The E and
od (matching each storm’s erosivity). The simulated soil L adjustments are well founded. The rationale for the B
loss in Fig. 3 does not include any adjustments for the lower adjustment is not clear. Should we expect runoff generated
kinetic energy of the simulator, or the differences in slope by a rainfall simulator to be delayed compared with natural
length. It represents the “raw” data from the low-intensity rainfall? It could be argued that the immediate imposition of
simulations. Although there was a significant correlation highly erosive rainfall on dry soils is unrelated to the pattern
(P < 0.10) between natural and simulated soil loss for the of natural rainstorms, where there is some gradual soil wet-
NOLAN ET AL. — STORM EROSION WITH A RAINFALL SIMULATOR 673

Table 3. Soil loss from natural rainfall on 144 m2 plots for CT, RT and ZT conditions during measurement periods from 1987 to 1991
Soil Loss (kg ha–1)
CT RT ZT
Measurement
period Mean SEz Mean SE Mean SE
1 69.3 20.8 31.3 11.5 41.0 3.5
2 67.3 18.2 35.3 15.8 6.7 2.0
3 487.5 146.0 233.6 36.0 40.2 12.0
4 406.8 42.8 277.2 7.5 208.0 26.0
5 71.6 14.4 66.2 21.1 34.9 2.3
Sum 1102.5a 126.0 643.6b 42.4 331.7c 33.0
zStandarderror.
a,bValues followed by different letters in the sum row are significantly different at P < 0.05 using analysis of variance.
Can. J. Soil. Sci. Downloaded from cdnsciencepub.com by 130.105.110.3 on 09/04/22
For personal use only.

Fig. 2. Cumulative soil loss from the GRS II for 60 mm h–1 and 140 mm h–1 simulations in CT, RT and ZT conditions. Bars represent stan-
dard error. Simulated soil loss is determined by matching the erosivity of each natural storm with a simulation duration that represents the
corresponding simulator erosivity.

ting before highly erosive periods. And should we also sented by the field conditions during our simulations in May
expect the runoff lag to increase with decreasing simulator 1991; ii) the crudeness of “scaling-up” simulator measure-
plot size, as described by Auerswald et al. (1992)? It is not ments to the larger area of the natural rainfall plots; iii) inac-
clear to us what the physical basis for this would be. curacy in R as a measure of erosivity; and iv) not enough
When we conclude that the low-intensity simulations replications in either the natural or simulated plots.
most accurately describe natural soil loss (after E-L-B In our methodology there is the potential for significant
adjustments), we dismiss the good agreement between nat- error when scaling up the simulator measurements. The L
ural and simulated soil loss from the high-intensity simula- factor is a simplistic treatment of slope length effects on soil
tion for CT and RT (unadjusted). We believe this agreement loss. For instance, the value of β in Eq. 8 (used to calculate
was coincidental given the well-founded expectation that α) implies equal amounts of sheet and rill erosion for a 22.1
soil loss from a 1-m length plot would be less than from a m slope. However, McCool et al. (1989) list α values rang-
30-m length plot due to the absence of rill erosion (which ing from 0.30 to 0.62 for a 7% slope, reflecting differences
was not so in the high intensity case), and the very poor in the proportion of rill to sheet erosion. When we vary α
agreement for ZT in the high-intensity simulation. over the range listed by McCool et al. (1989) we see a
Although we found good results using the E-L-B adjusted departure in our L adjustment factor from –27% to +43%,
low intensity simulation to estimate the accumulated natur- and a corresponding variation in our simulated soil loss.
al soil loss over the five study periods, the relationship was Inaccuracy in R is evident in the differences between soil
not as impressive on an individual measurement period loss for the high- and low-intensity simulations. If R accu-
basis (Fig. 5). We saw an underestimation of the high natur- rately quantified erosivity, intensity should not have made a
al soil losses, an overestimation of the low losses, and a difference in the simulated soil loss, as intensity is incorpo-
large amount of scatter in the relationship (r2 = 0.32). We rated in the calculation. Yet intensity did make a large dif-
attribute this scatter to four possible factors (excluding mea- ference. Perhaps our high simulation intensity (140 mm h–1)
surement errors): i) the field conditions that spanned 5 yr of was outside the range of rainfall conditions used to develop
spring conditions in our study were not universally repre- R, and therefore outside the range of R applicability. Also,
674 CANADIAN JOURNAL OF SOIL SCIENCE
Can. J. Soil. Sci. Downloaded from cdnsciencepub.com by 130.105.110.3 on 09/04/22

Fig. 3. Natural versus simulated soil loss for five measurement Fig. 5. Natural versus simulated soil loss for five measurement
periods and three replications of CT, RT and ZT conditions. periods in CT, RT and ZT conditions. Simulated soil loss is from
Simulated soil loss is from 60 mm h–1 intensity with no adjust- 60 mm h–1 intensity with E-L-B adjustments.
ments.
For personal use only.

Fig. 6. Soil loss from natural rainfall on 144 m2 plots versus R from
CT, RT and ZT conditions. Soil loss and R values are for five mea-
surement periods. Lines represent best linear fit regressions.

setting I30 to twice the rainfall depth when calculating Rsim,


as instructed by Wischmeier and Smith (1978) for rainfall
durations less than 30 min, may not be well founded.
The natural soil loss measurements also indicate deficien-
cies in R. Figure 6 shows how natural soil loss for the five
measurement periods varied with the accumulated period R
Fig. 4. Accumulated soil loss from five measurement periods in values. While we expected a close linear relationship
CT, RT and ZT conditions. The different columns represent soil between the two, we instead saw substantial scatter (r2 =
loss from: natural rainfall on 144 m2 plots (Natural); unadjusted
simulations (Unadjusted); energy adjusted simulations (E Adj.);
0.27 for CT, 0.17 for RT and 0.01 for ZT). While this may
energy plus slope length adjusted simulations (E-L Adj.); and ener- be partially explained by differences in field conditions at
gy, slope length and runoff begin adjusted simulations (E-L-B the time of the different storms, or possible inaccuracies in
Adj.). The two graphs represent low (60 mm h–1) and high (140 the precipitation record, some of this scatter is probably due
mm h–1) intensity simulations. Bars represent standard error. to inaccuracy in R. This would not be surprising, as R is an
NOLAN ET AL. — STORM EROSION WITH A RAINFALL SIMULATOR 675

empirical factor that relates common meteorological vari- range of rainfall conditions used in its development, and not
ables to erosivity, with inherent uncertainties. Indeed, use extreme rainfall intensities in simulations. We also rec-
Wischmeier and Smith’s (1958) original data show less ommend using a simulation intensity that is close to the
accuracy in the relationship between R and soil loss for the intensities of naturally occurring erosion events at a particu-
low values of R typical of our study. We might also specu- lar location (as was our low intensity simulation), and a
late that the poor relationship between R and natural soil duration that spans the R climatology of interest.
loss from ZT indicates inadequacy in R. Since the soil sur-
face is almost completely covered with crop residue in ZT, ACKNOWLEDGEMENTS
the impact energy of raindrops may no longer control soil We appreciate contributions to this study from K. Skarberg,
loss, and erosivity may be better related to another factor, G. Lohstraeter, K. Toogood, L. Goonewardene, K. Carroll,
such as volume of water applied. R. Kline, A. Koppel, P. Dzkowski, R. C. Izaurralde, R. G.
Finally, the fact that we saw good agreement between Kachanoski, and the reviewers. Financial support was
accumulations of natural and simulated soil loss, but not for received from the Canada-Alberta Soil Conservation
individual measurement periods, is not necessarily due to Initiative and the Canada-British Columbia Agri-Food
Can. J. Soil. Sci. Downloaded from cdnsciencepub.com by 130.105.110.3 on 09/04/22

shortcomings in our method. Wendt et al. (1986) found Regional Development Subsidiary Agreement.
large and unexplained variability in soil loss from uniform
plots for single rainstorms, but differences between plots Andraski , B. J., Daniel, T. C., Lowery, B. and Mueller, D. H.
1985. Runoff results from natural and simulated rainfall for four
were reduced when soil loss was summed over several
tillage systems. Trans. Am. Soc. Agric. Eng. 28: 1219–1225.
events. They suggest that a large number of plots are need- Auerswald, K., Kainz, M., Wolfgarten, H.-J. and Botscheck, J.
ed to accurately measure soil loss from single events. 1992. Comparison of German and Swiss rainfall simulators —
Perhaps the three repetitions used for measuring natural soil influence of plot dimensions. Z. Pflanzenernahr. Bodenk. 155:
loss in our study were not enough to give soil loss over a 493–497.
limited number of storms. However, it then follows that Blough, R. F., Jarrett, A. R., Hamlett, J. M. and Shaw, M. D.
three repetitions of simulated soil loss for each treatment 1990. Runoff and erosion rates from slit, conventional, and chisel
may not be enough either. tillage under simulated rainfall. Trans. Am. Soc. Agric. Eng. 33:
1557–1562.
For personal use only.

Elliot, W. J., Liebenow, A. M., Laflen, J. M. and Kohl, K. D.


SUMMARY AND CONCLUSIONS 1989. A compendium of soil erodibility data from WEPP cropland
We have shown that soil loss on large plots from a series of soil field erodibility experiments 1987 & 88. NSERL Report No.
natural rainstorms can be accurately estimated using soil 3, The Ohio State University & USDA Agricultural Research
loss measured with a rainfall simulator. Our method relied Service.
Foster, G. R., Meyer, L. D. and Onstad, C. A. 1977. A runoff
on matching the erosivity of the simulator to that of each
erosivity factor and variable slope length exponents for soil loss
rainstorm, using Wischmeier and Smith’s (1978) R factor to estimates. Trans. Am. Soc. Agric. Eng. 20: 683–687.
quantify erosivity. The best agreement occurred with a sim- Foster, G. R., McCool, D. K, Renard, K. G. and Moldenhauer,
ulation intensity of 60 mm h–1 after adjustments for: the W. C. 1981a. Conversion of the universal soil loss equation to SI
lower kinetic energy of the simulator compared with natur- metric units. J. Soil Water Conserv. 36: 355–359.
al rain, the smaller slope length of the simulator, and calcu- Foster, G. R., Simanton, J. R., Renard, K. G., Lane, L. J. and
lating erosivity from runoff start times. After these Osborne, H. B. 1981b. Discussion of “Application of the
adjustments, our simulator estimates of soil loss were with- Universal Soil Loss Equation to Rangelands on a Per-Storm
Basis,” by Trieste and Gifford in Journal of Range Management.
in 12% of the natural soil loss over three tillage conditions. 33: 66–70, 1980. J. Range Manage. 34: 161–165.
These results were surprisingly good given: i) that one 20- McCool, D. K., Foster, G. R., Mutchler, C. K. and Meyer, L. D.
min simulation was used to estimate 5 yr of springtime soil 1989. Revised slope length factor for the Universal Soil Loss
loss; ii) the smaller size of the simulator plots and the diffi- Equation. Trans. Am. Soc. Agric. Eng. 32: 1571–1576.
culties with “scaling-up” to larger areas; and iii) the weak- McIsaac, G. F. and Mitchell, J. K. 1992. Temporal variation in
ness of R as a quantification of erosivity. Our results were runoff and soil loss from simulated rainfall on corn and soybeans.
not as good for individual measurement periods, but this Trans. Am. Soc. Agric. Eng. 35: 465–472.
Nolan, S. C., Goddard, T. W. and van Vliet, L. J. P. 1992. Soil
may be due to the natural variability of soil loss from single
erosion from selected conservation and conventional management
erosion events. practices. Pages 90–103 in Proceedings of the 29th Alberta Soil
Our results suggest that a rainfall simulator can be used to Science Workshop. 18–20 Feb. Lethbridge, AB.
estimate natural soil loss when operated on the principle of Peterson, A. E. and Bubenzer, G. D. 1986. Intake rate: Sprinkler
Wischmeier and Smith’s (1978) R factor. Subsampling soil infiltrometer. Pages 845–867 in A. Klute, ed. Methods of soil
loss during rainfall simulations is required to construct a analysis. Part 1. Physical and mineralogical methods. Agronomy
cumulative soil loss curve which makes it possible to use Monogr. no. 9. 2nd ed. ASA, Madison, WI.
one simulator run (preferably repeated) to approximate any Robichaud, P. R. and Waldrop, T. A. 1994. A comparison of
surface runoff and sediment yields from low- and high-severity site
rainstorm (e.g., any R value). The cumulative soil loss curve preparation burns. Water Resources Bulletin. 30: 27–34.
is also essential for making adjustments to the reduced Romkens, M. J. M. and Wang, J. Y. 1986. Effect of tillage on
kinetic energy of the simulated rainfall. Because R is an surface roughness. Trans. Am. Soc. Agric. Eng. 29: 429–433.
empirical erosivity measure, we advise staying within the Tossell, R. W., Dickinson, W. T., Rudra, R. P. and Wall, G. J.
676 CANADIAN JOURNAL OF SOIL SCIENCE

1987. A portable rainfall simulator. Can. Agric. Eng. 29: 155–162. Wischmeier, W. H. and Mannering, J. V. 1969. Relation of soil
Tossell, R. W., Wall, G. J., Rudra, R. P., Dickinson, W. T. and properties to its erodibility. Proc. Soil Sci. Soc. Am. J. 33:
Groenevelt, P. H. 1990. The Guelph rainfall simulator II: Part 2. 131–137.
Comparison of natural and simulated rainfall characteristics. Can. Wischmeier, W. H. and Smith, D. D. 1958. Rainfall energy and
Agric. Eng. 32: 215–223. its relationship to soil loss. Trans. Am. Geophys. Union. 39:
van Vliet, L. J. P., Kline, R. and Hall, J. W. 1993. Effects of 285–291.
three tillage treatments on seasonal runoff and soil loss in the Peace Wischmeier, W. H. and Smith, D. D. 1978. Predicting rainfall
River region. Can. J. Soil Sci. 73: 469–480. erosion losses — a guide to conservation planning. Agric.
Wendt, R. C., Alberts, E. E. and Hjelmfelt, Jr., A. T. 1986. Handbook No. 537. USDA, Washington, DC.
Variability of runoff and soil loss from fallow experimental plots.
Soil Sci. Soc. Am. J. 50: 730–736.
West, L. T., Miller, W. P., Langdale, G. W., Bruce, R. R.,
Laflen, J. M. and Thomas, A. W. 1991. Cropping system effects
on interrill soil loss in the Georgia Piedmont. Soil Sci. Soc. Am. J.
55: 460–466.
Can. J. Soil. Sci. Downloaded from cdnsciencepub.com by 130.105.110.3 on 09/04/22
For personal use only.

You might also like