Professional Documents
Culture Documents
Kin Nell 2000
Kin Nell 2000
Kin Nell 2000
www.elsevier.com/locate/envsoft
Abstract
In the Agricultural Non Source Pollution model (AGNPS), grid-cell erosion is predicted for individual rainfall events using the
Universal Soil Loss Equation (USLE) even though the USLE is not well suited for this purpose. A new modification of the USLE,
the USLE-M, is better suited to predicting event erosion and also provides a mechanism for accounting for the impact of upslope
runoff on erosion that is not available with the USLE. A software system that replaces the USLE by the USLE-M in AGNPS is
described and its impact illustrated by an example based on a 1-in-10 year event on a 2300 ha catchment near Nundle, New South
Wales, Australia. 2000 Elsevier Science Ltd. All rights reserved.
1364-8152/00/$ - see front matter 2000 Elsevier Science Ltd. All rights reserved.
PII: S 1 3 6 4 - 8 1 5 2 ( 0 0 ) 0 0 0 0 2 - 5
332 P.I.A. Kinnell / Environmental Modelling & Software 15 (2000) 331–341
scale. Data and computer constraints mean that, for prac- in the USLE is replaced by a runoff rate factor because
tical purposes, the USLE will continue to provide the watershed characteristics such as drainage area, stream
basis for modelling rainfall erosion in catchments for slope, and watershed shape influence runoff rates and
some time to come. delivery ratios in a similar manner. As a consequence of
The USLE can be written as this, Williams proposed an equation that can be writ-
ten as
A⫽R K L S C P (1)
SYe⫽XeK L S CePe (5)
where A is the average annual soil loss (mass/area/year),
R is the erosivity index determined by where SYe is the event sediment yield,
Xe⫽a(Qe qp)0.56
冘
N (6)
(EI30)n where a is an empirical coefficient, Qe is runoff amount
n⫽1
R⫽ (2) and qp is the peak runoff rate obtained during the erosion
Y event, and K, L, S, Ce, Pe as defined for the USLE. This
where E is the total kinetic energy of a rainstorm, I30 is model has become known as the Modified Universal Soil
the maximum 30 min intensity, Y is the number of years Loss Equation (MUSLE). In a comparison on 11 3-acre
during which N rainfall events occurred, K is the soil (1.2 ha) watersheds where delivery ratios were con-
factor, L is the factor accounting for the influence of sidered to be 1.0, Williams observed that, with the Ce,
slope length, S is the factor accounting for the influence and Pe set to their average annual values (C, P), Eq. (5)
of slope gradient, C is the crop and crop management explained 82% of the variation while Eq. (3) explained
factor and P is the conservation protection factor. 47%. However, Foster et al. (1982) noted that a major
Although the USLE is not designed to predict event advantage of an erosivity index that includes runoff
erosion, the event version of it can be written as terms is the elimination of large overestimates of soil
loss when runoff is negligible and rainfall amounts and
Ae⫽ReKeL S CePe (3) rates are great. The advantage of the Williams index over
where Ae is the soil loss for an event, the USLE is probably more related to the failure of the
USLE to consider runoff explicitly as a factor in
Re⫽EI30 (4) determining Re than anything else.
and Ke, Ce, Pe are factors accounting for the soil, crop In MUSLE, Re in the USLE, the EI30 index, is
and crop management, and conservation protection con- replaced by a (Qe qp)0.56 while all the other parameters
ditions that exist at the time of the event. remain as defined for the USLE. In the EPIC model
Soil erosion results from soil being detached from the (originally the Erosion/Productivity Impact Calculator
soil surface and the subsequent transport of the detached but now Environmental Policy Integrated Climate), a
soil away from the point of detachment. The sediment continuous simulation model developed by Williams et
delivered from land surface to a water course (river, al. (1984),
stream, gully) or impoundment (dam, lake) is dependent SYe⫽XeK L S CePeROKF (7)
on gross erosion and all the processes which affect deliv-
where ROKF is the coarse fragment factor as defined
ery from the point of detachment to the water course or
by Simanton et al. (1984), Xe is selected from one of
impoundment. By convention, units of mass per unit area
the following:
are used for amounts of soil erosion. For runoff and soil
loss plots such as those used to evaluate parameters in Xe⫽EI30 (8a)
the USLE, the product of erosion amount and area gives Xe⫽1.586(Qe qp) 0.56
DA 0.12
(8b)
the sediment yield in units of mass. Runoff and soil loss
plots are small watersheds or catchments. However, as Xe⫽0.65EI30⫹0.45(Qe qp) 0.33
(8c)
catchment size increases, areas of deposition within the where DA is drainage area expressed in ha, Qe is
catchment tend to reduce the sediment yield below that expressed in mm, qp in mm/h, EI30 in MJ.mm/ha.h and
predicted from erosion models like the USLE. Under SYe in t/ha (Williams and Arnold, 1997). Values of K,
these circumstances, a delivery ratio is used to convert L, S, Ce, Pe and ROKF are determined independently of
estimates of gross erosion to sediment yield (Williams the index used for Xe. However, in the USLE, K is, for
et al., 1971). The sediment delivery ratio is the ratio of example, calculated by
冘
the sediment yield at a specific location in the watershed N
and the gross erosion upstream of that point. While a (Ae)n
sediment delivery ratio is considered necessary to deter- n⫽1
K⫽ (9)
冘
mine sediment delivery from erosion estimated using the N
USLE in catchments, Williams (1975) contended that the (EI30)n
delivery ratio is not necessary if the rainfall energy factor n⫽1
P.I.A. Kinnell / Environmental Modelling & Software 15 (2000) 331–341 333
when L = S = C = P =1.0. Thus it follows that, for EPIC, manner in which data are stored and retrieved in modern
Geographic Information Systems (GIS) facilitates the
冘
N
冘
T
mary cell–subcell approach, it is more appropriate to use
small grid cells so that the number of cells allocated to Re⫽ (QEA)t (11)
t⫽1
a hydrologic unit depends only on the size of the
hydrologic unit and the size of the grid cells used. The where T is the number of time units in a rainfall event,
334 P.I.A. Kinnell / Environmental Modelling & Software 15 (2000) 331–341
and Q is the runoff and EA is the rainfall energy in each when L = S = PUM =1.0 and C ⫽1.0. A similar
time unit, could be used as an alternative to EI30 and expression is used to determine PUM. As with the USLE,
that the excess rainfall rate (Ix, the difference between CUM =1.0 for bare fallow, and PUM =1.0 for cultivation
the rainfall rate and the infiltration rate) could be used up and down the slope. Since variations in slope length
as a surrogate for Q. This approach assumes that short have, in theory, no appreciable impact on runoff, and
term variations in sediment concentration (c) are directly S values have been developed through observation that
related to EA. One difficulty associated with the use of variations in slope gradient have no significant impact on
Eq. (11) is that adequate knowledge of the short term runoff when L is held constant, arguably, the topographic
variations values of Q (or Ix) and EA during a rainfall factors used in the USLE or the Revised USLE (RUSLE)
event is often lacking. However, Kinnell (1997) noted (Renard et al., 1997) still apply to USLE-M.
that Fig. 1 provides a comparison of the abilities of the
USLE and USLE-M to account for soil loss from a bare
Ae⫽Qecbe (12)
fallow runoff and soil loss plot in the USA. The logarith-
where mic form of the Nash-Sutcliffe model efficiency statistic,
冘 冘
T N
qtct (log YO−log YC)2n
cbe⫽
t⫽1 n⫽1
(13) Z(log)⫽1⫺ (20)
Qe
冘
N
(log YO−log Y ) 2
M n
and that cbe was dependent on the kinetic energy per unit n⫽1
quantity of rain (= E/Be where Be is event rainfall
amount) and a measure of the peak rainfall intensity where YO is the measured value, YC the computed value
since the peak rainfall intensity tends to produce the and YM the mean of the measured values, is shown as
highest sediment concentration and the highest runoff an indicator of model performance. This version of the
rate during a rainfall event. Kinnell observed that if I30 statistic is used here because it is influenced more by a
was assumed to provide an adequate measure of this model’s performance with respect to small to medium
intensity, erosion events than the normal version. Fig. 2 shows the
Z(log) values for the two models plotted against the
Re⫽QeI30E/Be (14) Gross Infiltration Ratios for runoff producing events
and since the runoff ratio (QR) is given by, (GIRrope) occurring on the plots at the 14 locations in the
USA and one in Australia. The GIRrope value for each
QR⫽Qe/Be (15) plot was calculated by
冘
Eq. (14) can be written as N
Ae⫽RUMeKUMeL S CUMePUMe (17) for only those rainfall events that produced runoff over
the period of observation (usually 4 or more years). GIR-
where RUMe = QR EI30, and the subscript UM indicates
rope generally has a lower value than the usual Gross
a parameter with values that differ from the USLE. It Infiltration Ratio which is based on all rainfall events
follows from Eq. (10) that irrespective of whether they produced runoff or not. As
冘
N
2). The data presented in Fig. 2 indicate that the USLE
(Ae)n operates best when soils are impervious or close to
impervious and that, as a general rule, USLE-M will pre-
CUM⫽
n⫽1
(19)
冘
N dict event soil loss better than the USLE if runoff can
KUM (QREI30)n be estimated adequately. This result is consistent with
n⫽1 the observation of Foster et al. (1982) that a major
P.I.A. Kinnell / Environmental Modelling & Software 15 (2000) 331–341 335
Fig. 1. Relationships between event soil losses for plot 8 in experiment 1 at Arnot (Ithaca), NY and the EI30 and QREI30 indices. The lines
represent the relationships generated by: (A) the USLE; and (B) USLE-M. From Kinnell and Risse (1998).
advantage of an erosivity index that includes runoff divided into a number of segments which could be
terms with rainfall terms in a product is the reduction of assumed to have uniform slope gradient and soil proper-
large overestimates of soil loss when runoff is negligible ties. As a result of this, they developed an equation for
and rainfall amount and rates are great. Thus, for pre- calculating the L-factor for the ith segment:
dicting event erosion such as in AGNPS, USLE-M is, lm+1 −lm+1
Li⫽
i i−1
in theory, a better model to use than the USLE. (23)
(li−li−1)(22.13)m
1.2.1. The L factor when the USLE-M is applied to where li is the distance from the upslope boundary of
grid cells the field or hillslope to the lower boundary of the ith
As noted earlier, Desmet and Govers (1996) segment. Desmet and Govers (1996) extended Eq. (23)
developed a method of determining L in grid cells that to areas where square elements (grid cells) are assumed
are not separate hydrologic units. The L factor for a slope to have uniform slope gradient and soil properties. For
is given by the grid cell with co-ordinates i, j, Desmet and Govers
consider the L-factor to be described by
L⫽(l/22.13)m (22)
(Ai,j-in+D2)m+1−Am+1
Li,j⫽ m+2 m
i,j-in
where l is the length of the slope as measured along the (24)
horizontal projection and m varies with factors such as D xi,j(22.13)m
slope gradient (Wischmeier and Smith, 1965). Foster and where Ai,j-in is the area contributing to flow into the cell
Wischmeier (1974) considered that the slope could be with co-ordinates i, j, D is the length of the sides of the
336 P.I.A. Kinnell / Environmental Modelling & Software 15 (2000) 331–341
Fig. 2. The relationships between Z(log) for the USLE and USLE-M and the gross infiltration ratio for runoff producing events (GIRrope) for bare
fallow plots at the 14 USA locations and Gunnedah in Australia. Z(log) values for USLE-M are represented by triangles, those for the USLE by
circles. The relationship between Z(log) and GIRrope for USLE-M is indicated by the solid line, that for the USLE by the dashed line. From Kinnell
and Risse (1998).
grid cell, and xi,j is the width of the contour over which where QCe.i,j = Qe.i,j/Be and QCe.i,j-in = Qe.i,j-in/Be. It should
the flow is discharged. xi,j is dependent on flow direction be noted that QRe.i,j-cell is the ratio of the runoff volume
relative to grid cell orientation. from the cell divided by the volume of rain that falls on
The concept behind Eqs. (23) and (24) is that the ero- the area of the cell. Since runoff from upslope contrib-
sion in an element or grid cell can be determined by utes to the volume of runoff from the cell, QRe.i,j-cell can
subtracting the sediment discharged (erosion per unit take on values greater than 1.0. In contrast, QCe.i,j and
area multiplied by area) for the area upslope of the QCe.i,j-in normally have values that are less than 1.0.
element or grid cell from the sediment discharged for There are two extremes to consider with respect to
the area that includes the cell and dividing the result by Eq. (29). The first is when the whole of the eroding area
the area of the segment or cell. If this concept is applied is impervious. In this case, QCe.i,j = QCe.i,j-in =1 and
when the USLE-M is used, and if Qe.i,j is the runoff (in LUM.e.i,j equals Li,j as calculated by Eq. (24). The other
units of depth) passing across the lower boundary of the extreme is when no runoff enters the cell from upslope.
cell i, j and Qe.i,j-in is the runoff passing across the upper Under these circumstances,
boundary of the cell during an event, then
QCe.i,j(Ai,j-in+D2)m+1
Qe.i,j(Ai,j-in+D ) −Qe.i,j-inA
2 m+1 m+1 LUMe.i,j⫽ (30)
L⬘UM.e.i,j⫽
i,j-in
(25) QRe.i,j-cellDm+2xmi,j(22.13)
m
BeDm+2xm i,j(22.13)
m
while it follows from Eqs. (17) and (22) that
when
LUMe.i,j⫽(D/22.13)m (31)
Ae.i,j⫽EI30K L⬘
S CUMe.i,jPUMe.i,j
UMe.i,j UMe.i,j i,j (26)
However, the values of LUMe.i,j generated by these two
equations are not equal and the difference between them
Because the erosivity index for a cell when the USLE- increases as the number of cells in the upslope contribu-
M is used is given by the product of E, I30 and the runoff ting area increases. The discrepancy is eliminated by
ratio for the cell (QRe.i,j-cell), (Kinnell, in review)
Ae.i,j⫽[QRe.i,jEI30]KUMe.i,jLUMe.i,jSi,jCUMe.i,jPUMe.i,j (27) QCe.i,j(Ai,j-in+D2)m+1−QCe.i,j-inAm+1
LUMe.i,j⫽F
i,j-in
(32)
so that QRe.i,j-cellDm+2xm
i,j(22.13)
m
LUMe.i,j values determined by Eq. (32) are hydrodyn- network has not been otherwise mapped. User online
amic and vary between erosion events in contrast to the input is required to generate the attribute files for running
values of Li,j determined by Eq. (24) which do not TOPAZ and Program B. For TOPAZ, the location of the
change between events. A comparison of the use of the catchment outlet is a required input in addition to cell
USLE with Li,j values determined by Eq. (24) and the size. The initiation points for the channel network are
USLE-M with LUMe.i,j determined by Eq. (32) on the pre- determined from a critical source area (CSA). Attributes
diction of erosion in grid cells associated with a sub- associated with soils and land use such as Curve Num-
catchment in Queensland, Australia, is given in Kinnell ber, Ke, KUMe, Ce and CUMe are required for Program B.
(in review). AGNPS 5.00 uses traditional units for K (T.A.h/100.A-
ft.T.in, ton.acre.hour/100.acre-foot.ton.inch) and EI30
(ft.T.in/A.h). L, S, C, and P have no units. AGNPS v5.00
2. Applying the USLE-M within AGNPS is run on one of the AGNPS input files (the standard
AGNPS input) and the output used together with the
AGNPS software is public domain and readily avail- other AGNPS input file (the modified AGNPS input file
able from sources such as the Internet. Until recently, #1) to produce a second modified AGNPS input file that,
AGNPS used the USLE to predict cell erosion. A new when used with AGNPS v5.00, produces the AGNPS-
version (AGNPS98) that uses the Revised Universal Soil UM output file. The standard AGNPS output file pro-
Loss Equation (RUSLE) was released in 1998. The duced by the first use of AGNPS v5.00 provides the
software that has been developed to apply the USLE-M hydrologic data required to determine the runoff coef-
has been designed for use with AGNPS version 5.00 on ficients (QCe.i,j, QCe.i,j -in) and runoff ratios (QRe.i,j-cell)
a PC running under DOS or a DOS window under Win- required by the USLE-M. The software has a limit of
dows 95. 32 000 cells, this limit being the limit set by AGNPS
The primary data inputs are contained in 3 GIS Ascii v5.00.
files. (Fig. 3). The GIS files for elevation, soils and lan- As noted earlier, in AGNPS, a restriction of 1000 ft
duse can have the formats of either ARC/INFO or (305 m) is placed on slope length within a primary cell
GRASS Ascii input/output files. Many GIS programs on the grounds that deposition or channelization can be
can write these formats so that the software is largely expected to occur within this distance. However, in
GIS independent. A dedicated program (Program A) many catchments, slope lengths actually exceed this
uses these files and, in association with online input from length. The AGNPS-UM software does not place a
the user, produces files that are used by Program B to restriction on length of slope associated with a set of
generate two AGNPS input data files. A public domain contiguous grid cells but relies on overland flow path
program, TOPAZ, is used to generate the files containing lengths determined by the channel network produced by
cell slope gradients and flow directions required by Pro- either TOPAZ or directly by the user. Consequently, the
gram B. TOPAZ also produces a synthesised channel user needs to ensure that the channel network used is
network that can be used by Program B if a channel appropriate. TOPAZ provides a capacity to specify dif-
Fig. 3. Schematic of the software system for replacing the USLE by the USLE-M in AGNPS v5.00.
338 P.I.A. Kinnell / Environmental Modelling & Software 15 (2000) 331–341
ferent CSA values for different parts of the catchment soils were 0.40 T.A.h/100.A-ft.T for soil 1, 0.60
if required. With respect to deposition, that is an issue T.A.h/100.A-ft.T for soil 2 and 0.47 T.A.h/100.A-ft.T
dealt with by the sediment transport model in AGNPS for soil 3. These values are associated with GIRrope
when it is applied to the grid cells. values of 0.75 (soil 1) and 0.63 (soils 2 and 3). Soil 1
An example illustrating the impact of using the USLE- is the same type of soil as that associated with the bare
M to model cell erosion within AGNPS is presented here fallow plots at Gunnedah used in the USLE — USLE-
using a 1-in-10 year 6 hour storm on the 2300 ha Back M comparison shown in Fig. 2.
Creek Catchment near Nundle, New South Wales, Aus- Fig. 5 provides a comparison between the erosion pre-
tralia. The outlet of the catchment is at the north east dicted for the 1-in-10 year 6 hour event by AGNPS
corner of the catchment. Slope gradients in excess of v5.00 with the USLE and L predicted using the Desmet
30% occur quite frequently in the southern part (Fig. 4). and Govers method (AGNPS-DG) and with the USLE-
The channel network shown in Fig. 4 was produced by M and L predicted using Eq. (32) (AGNPS-UM) when
TOPAZ using a CSA value of 15 ha. The network erosion in the 100 m by 100 m (1 ha) cells is considered
matched the channel network observed in ariel photo- in units of t/ha (metric ton per hectare) rather than T/A
graphs reasonably well but not sufficiently well for the (ton per acre) as is customarily used in the US
example to be considered anything more than a demon- (conversion: 2.242 t/ha =1 T/A). The 1-in-10 year 6 hour
stration. event produces 2.8 inch rain and an EI30 value of 62
The catchment contains 3 different soils (Fig. 4). A ft.T.in/A.h. The runoff coefficients associated with the
clay soil with a USLE K value of 0.30 T.A.h/100.A-
event varied from 0.09 to 0.28 while QRe.i,j-cell ranged
ft.T.in (soil 1) dominates the southern half of the catch-
from 0.09 to 4.02 in the cells with overland flow when
ment. A clay soil with a USLE K value of 0.38
the critical source area for channel initiation was 15 ha
T.A.h/100.A-ft.T (soil 2) dominates the northern half
and overland flow path lengths of up to about 600 m
while an alluvial soil (soil 3) with a USLE K value of
occur. LUMe.i,j ranged from 0.14 to 5.79. In comparison,
0.30 T.A.h/100.A-ft.T is found near the main drainage
channel in this northern area. The catchment is predomi- L factor values for the USLE determined by the Desmet
nantly grazing land with a small area cropped with wheat and Govers method ranged from 1.42 to 16.41. The aver-
in the north eastern part of the catchment. The 1-in-10 age cell erosion predicted by the USLE-M was 10.3 t/ha
year 6 hour storm is considered to occur in late July compared with 31.2 t/h for the USLE with L determined
when C for the wheat crop is 0.15 and the pasture had by the Desmet and Govers method. The sediment yield
been grazed to provide a cover of about 40% and a C at the outlet of the catchment predicted when the USLE-
value of 0.10. The CUM values for the pasture and wheat M was used was 1965 t as compared with 6173 t pre-
areas were 0.20 and 0.28 respectively. The soil was con- dicted using the USLE and the Desmet and Govers
sidered to be fairly dry at the onset of the event and, as method for predicting slope length. Obviously, had the
a result, 64 was used for the Curve Numbers (CN) for catchment been wetter at the onset of the event, then a
soils 1 (southern half) and 3 (alluvial) under pasture. CN higher proportion of the cells would have higher erosion
=72 and CN =74 were used for soil 2 under pasture and amounts predicted by the USLE-M leading to a higher
wheat respectively. The KUM values allocated to these sediment yield.
Fig. 4. Slope gradients (%), K (US units) and C values, and the channel network in the Back Creek catchment, near Nundle, New South Wales,
Australia. Grid cells are 100 m by 100 m (1 ha).
P.I.A. Kinnell / Environmental Modelling & Software 15 (2000) 331–341 339
Fig. 5. Cell erosion predicted by the USLE with L determined by the Desmet and Govers method (AGNPS-L via D and G) and the USLE-M
(AGNPS-UM) for a 1-in-10 event under the conditions given in the text. Grid cells are 100 m by 100 m (1 ha).
Fig. 6. Cell erosion predicted by the USLE with slope length restricted to the size of cell and the ratio of cell erosion predicted by the USLE-
M (AGNPS-UM) to cell erosion predicted by the USLE with slope length restricted to the size of cell (AGNPS- cell L) for the 1-in-10 year event.
Grids cell are 100m by 100m (1 ha).
4. Summary and conclusion rainfall-runoff erosivity factors for individual storms. Trans. ASAE
25, 124–129.
Foster, G.R., Wischmeier, W.H., 1974. Evaluating irregular slopes for
AGNPS is a grid-cell based model whose purpose is soil loss prediction. Trans. ASAE 17, 305–309.
to predict the spatial variation of rainfall erosion in He, C., Riggs, J.F., Kang, Y.T., 1993. Integration of geographic infor-
catchments and the impact of that erosion on water qual- mation systems and computer model to evaluate agricultural runoff
ity. It uses the USLE together with a sediment transport on water quality. Water Resources Bulletin 29, 891–900.
model to achieve this purpose. Unfortunately, the USLE Kinnell, P.I.A., 1995. The IxEA index: An index with the capacity to
give more direct consideration of hydrology in the USLE modelling
is not designed to predict event erosion well and the pro- environment. J. Soil Water Cons. 50, 507–512.
cedures available to determine the slope length factor for Kinnell, P.I.A., 1997. Runoff ratio as a factor in the empirical model-
a cell does not enable the impact of variations in upslope ling of soil erosion by individual rainstorms. Aust. J. Soil Res. 35,
runoff on cell erosion to be determined adequately. In 1–13.
theory, a new modification of the USLE called the Kinnell, P.I.A., Risse, L.M., 1998. USLE-M: Empirical modelling rain-
fall erosion through runoff and sediment concentration. Soil Sci.
USLE-M (Kinnell and Risse, 1998) provides a mech- Soc. Amer. J. 62, 1667–1672.
anism for overcoming these problems. A software sys- Kinnell, P. I. A. (in review). Slope-length factor for applying the
tem has been developed that replaces the USLE by USLE-M to erosion in grid cells. Soil & Tillage Research.
USLE-M in AGNPS v5.00. In theory, this software pro- Laflen, J.M., Lane, L.J., Foster, G.R., 1991. WEPP: A new generation
vides an improved modelling system for within catch- of erosion prediction technology. J. Soil and Water Cons. 46,
34–38.
ment erosion and the impact of that erosion on water Mitchell, J.K., Engel, B.A., Srinivasan, R., Wang, S.S.Y., 1993. Vali-
quality. dation of AGNPS for small watershedsusing integrated
AGNPS/GIS system. Water Resources Bulletin 29, 833–842.
Morgan, R.P.C., Quinton, J.N., Smith, R.E., Govers, G., Posen, J.W.A.,
Acknowledgements Auerswald, K., Chisci, G., Torri, D., Styczen, M.E., 1998. The Eur-
opean soil erosion model (EUROSEM): A dynamic approach for
predicting sediment transport from fields and small catchments.
The author wishes to acknowledge R. Beecham, Earth Surface Processes and Landforms 23, 527–544.
Department of Land and Water Conservation, New Renard, K.G., Foster, G.R., Weesies, G.A., McCool, D.A., Yoder,
South Wales, Australia for the GIS data used in this D.C., 1997. Predicting soil erosion by water: A guide to conser-
paper. vation planning with the Revised Universal Soil Loss Equation
(RUSLE). In: Agric. Handb. 703. US Government Print Office,
Washington, DC.
Simanton, J.R., Rawizt, E., Shirley, E.D., 1984. Effects of rock frag-
References ments on erosion of semi-arid rangeland soils. In: Erosion and
Productivity of Soils Containing Rock Fragments. Soil Sci. Soc.
Desmet, P.J.J., Govers, G., 1996. A GIS procedure for automatically Amer, Madison, WI, 65–67.
calculating the USLE LS factor on topographically complex land- Tim, U.S., Jolly, R., 1994. Evaluating agricultural non-point source
scape units. J. Soil and Water Cons. 51, 427–433. pollution using integrated geographic information system and
Foster, G.R., Lombardi, F., Moldenhauer, W.C., 1982. Evaluation of hydrologic/water quality model. J. Envirion. Qual. 23, 25–35.
P.I.A. Kinnell / Environmental Modelling & Software 15 (2000) 331–341 341
Williams, J.R., 1975. Sediment-yield prediction with universal equ- Wischmeier, W.H., Smith, D.D., 1965. Predicting rainfall erosion
ation using runoff energy factor. In: Present and Prospective Tech- losses from cropland east of the Rocky Mountains. In: Agric.
nology for Predicting Sediment Yield and Sources. ARS.S-40 US Handb. 282. US Government Print Office, Washington, DC.
Gov. Print Office, Washington, DC pp. 244–252. Wischmeier, W.H., Smith, D.D., 1978. Predicting rainfall erosion
Williams, J.R., Arnold, J.G., 1997. A system of erosion-sediment yield losses — A guide to conservation planning. In: Agric. Handb. 537.
models. Soil Technology 11, 43–55. US Government Print Office, Washington, DC.
Williams, J.R., Hiler, E.A., Baird, R.W., 1971. Predicting sediment Young, R.A., Onstad, C.A., Bosch, D.D., Anderson, W.P., 1987.
yields from small watersheds. Trans. ASAE 14, 1157–1162. AGNPS, Agricultural-Non-Point-Source Pollution model; A large
Williams, J.R., Jones, C.A., Dyke, P.T., 1984. The EPIC model and watershed analysis tool. In: Conservation Research Report 35.
its application. In: Proc Intl. Symp. on Minmum Data Sets for USDA-ARS, Washington, DC.
Agrotechnology Transfer, March 21–26, 1983. ICRISAT Center,
India.