Water erosion prediction at a national scale for South Africa

JJ Le Roux1*, TL Morgenthal2, J Malherbe2, DJ Pretorius3 and PD Sumner1

Department of Geography, Geoinformatics and Meteorology, University of Pretoria, Pretoria 0002, South Africa
1

2
Institute for Soil, Climate and Water, Agricultural Research Council, Private Bag X79, Pretoria 0001, South Africa
3
National Department of Agriculture, Directorate: Land Use and Soil Management, Private Bag X120, Pretoria 0001, South Africa

Abstract

Erosion is a major soil degradation problem in South Africa, confronting both land and water resource management through-
out the country. Given the increasing threat of soil erosion, a need to improve techniques of estimating the soil-erosion
risk at a national scale was identified by the National Department of Agriculture and forms the basic premise of this study.
Principles and components of the Revised Universal Soil Loss Equation are applied here since the model combines sufficient
simplicity for application on a national scale with a comprehensive incorporation of the main soil-erosion factors. Indica-
tors of erosion susceptibility of the physical environment, including climate erosivity, soil erodibility and topography were
improved over earlier assessments by feeding current available data into advanced algorithms. Two maps are presented: an
actual erosion-risk distribution, and a potential erosion-risk map that excludes the vegetation cover factor. Actual soil-erosion
risk, which relates to the current risk of erosion under contemporary vegetation and land-use conditions, was accounted for
by regression equations between vegetation cover and MODIS-derived spectral index. The area of land with a moderate to
severe potential risk is found to total approximately 61 m. ha (50%). Although more than 91 m. (75%) are classified as having
only a very low to low actual risk, approximately 26 m. ha (20%) of land is eroded at a rate greater than a soil-loss tolerance of
10 t/ha·yr, showing the potential to target erosion control to problem areas. The Eastern Cape, Limpopo and KwaZulu-Natal
Provinces have the highest erosion potential. Comparison of potential and actual erosion risk indicates that over 26 m. ha
(>30% of national land) could be subject to high erosion risk without maintenance or careful management of the current veg-
etation cover and land use. Although the distribution of the actual erosion risk broadly follows that outlined previously, this
study provides an advance on previous assessments of erosion; results are validated more comprehensively than before, and
show an overall accuracy of 77%. The paper also describes many of the limitations inherent in regional erosion studies.

Keywords: water erosion, national scale, potential risk, actual risk, RUSLE

Introduction delivery to rivers. Eroded soil material leads to sedimentation/

siltation of reservoirs, as well as an increase in pollution due
Soil erosion is an important form of land degradation and is to suspended sediment concentrations in streams which affects
among the world’s and South Africa’s most critical environmen- water use and ecosystem health (Flügel et al., 2003). Accord-
tal issues. Previous research indicates that more than 70% of ing to the latest State of Environment Report of SA, soil-erosion
South Africa (SA) is affected by varying intensities of soil ero- costs an estimated R2 bn. annually including off-site costs for
sion (Garland et al., 2000). Erosion is a process of detachment purification of silted dam water (Hoffman and Ashwell 2001;
and transportation of soil materials by wind or water (Morgan, cited in Gibson et al., 2006). Before prevention of soil erosion or
1995) and although 25% of SA is highly susceptible to wind ero- remediation can be undertaken, the spatial extent of the problem
sion (Hoffman and Todd, 2000), water is the dominant agent should be established.
causing erosion in SA and forms the focus of the study. Water Table 1 provides a summary of regional-based work
erosion occurs mostly through rain-splash, in un-concentrated undertaken on soil erosion in SA since 1990. Although some
flow as sheet erosion, as well as in concentrated flow as rill approaches are based on the collection of distributed field obser-
and/or gully erosion. Outcomes depend on the combined and vations and/or sediment data, most of the studies use a combina-
interactive effects of erosion factors, namely rainfall erosivity, tion of remote sensing and modelling techniques. In 1993, the
soil erodibility, slope steepness and slope length, crop manage- Agricultural Research Council – Institute for Soil, Climate and
ment, and support practice. More detail on the factors governing Water (ARC-ISCW) was contracted by the Department of Agri-
erosion, specifically in a South African context, is provided by culture (DoA) to investigate the use of remote sensing and GIS
Laker (2004). Although soil erosion is a natural process, it is in soil degradation management. As a result, Pretorius (1995)
often accelerated by human activities such as clearing of vegeta- produced the Erosion Susceptibility Map (ESM) at a scale of
tion or by overgrazing (Snyman, 1999). Loss of fertile topsoil 1:2.5 million by integrating a green vegetation cover map from
and reduction of soil productivity is coupled with serious off- NOAA satellite data with the sediment yield map of Southern
site impacts related to increased mobilisation of sediment and Africa (Rooseboom, 1992). Research continued in 1998 to
produce the Predicted Water Erosion Map (PWEM) at a scale
of 1:2.5 million applying the widely used Universal Soil Loss
* To whom all correspondence should be addressed. Equation (USLE) within a GIS framework (Pretorius, 1998).
Current address: Institute for Soil, Climate and Water, Agricultural
Methodology, however, is based on a considerable simplifica-
Research Council, Private Bag X79, Pretoria 0001, South Africa
 +2712 310-2684; fax: +2712 323-1157; tion of the USLE, by grouping some of the erosion factors (soil
e-mail: [email protected] and slope) as one. Furthermore, ESM and PWEM only provide
Received 18 December 2007; accepted in revised form 15 May 2008. percentage differences in erosion between regions without

Available on website http://www.wrc.org.za 305

ISSN 0378-4738 = Water SA Vol. 34 No. 3 July 2008
ISSN 1816-7950 = Water SA (on-line)
 Table 1
Summary table of regional erosion studies since 1990
Abbre-
Name Developed by Aim Area and scale
viation
GLASOD Global Assess- International Soil Actual soil erosion based on distributed point Global
ment of Human- Reference and Information data obtained from various experts. Soil-
Induced Soil Centre (ISRIC) erosion areas were delineated according to their Expert/subjective
Degradation (Oldeman et al., 1991) judgment. delineations
SDPM Sediment Water Research Commis- To provide spatial data on sediment yield by Southern Africa
Delivery sion (WRC) gathering sediment data and relevant geograph-
Potential Map (Rooseboom et al., 1992) ical information which influences sediment Catchments
yield values of catchments 14 to 60 000 km 2
BSI Bare Soil Index Agricultural Research To detect bare soil and the status of extensive South Africa
Council – Institute for Soil, eroded areas on a national scale with Landsat
Climate and Water (ARC- Thematic Mapper (TM) data. 30 m
ISCW)
(Pretorius and Bezuiden-
hout, 1994)
ESM Erosion Suscepti- Agricultural Research To investigate the use of remote sensing and
bility Map Council – Institute for Soil, GIS in soil degradation management by inte- South Africa
Climate and Water (ARC- grating a green vegetation cover map produced
ISCW) from NOAA AVHRR satellite data with the 1:2.5 million
(Pretorius, 1995) sediment yield map.
PWEM Predicted Water Agricultural Research Map erosion by integrating the main erosion
Erosion Map Council – Institute for Soil, contributing factors of the USLE in a GIS South Africa
Climate and Water (ARC- including the rainfall erosivity map of Smithen
ISCW) and Schulze (1982), the sediment yield map 1: 2.5 million
(Pretorius, 1998) and green vegetation cover map to account for
rainfall, soil-slope and vegetation factors.
NRA Natural Agricultural Research Map erosion by regional application of RUSLE Mpumalanga &
Resources Council – Institute for Soil, in a GIS. Soil and topography factors were, for Gauteng provinces
Auditing Climate and Water (ARC- the first time, separately facilitated by: Applica-
ISCW) tion of digital elevation models with a resolu- 1: 250 000
(Wessels et al., 2001a) tion of 75 m for the topography factor; and
(Wessels et al., 2001b) Soil maps (Soil Survey Staff, 1973-1987) were
used to link erodibility values to correspond-
ing soil series in the Land Type Inventories on
a scale of 1:250 000 (Land Type Survey Staff,
1972-2006).
ISRDS Integrated Sus- Agricultural Research As above OR Tambo and
nodes tainable Rural Council – Institute for Soil, Umkhanyakude
Development Climate and Water (ARC- nodes in Eastern
Strategy nodes ISCW) Cape and Kwa-
(Ströhmenger et al., 2004) Zulu-Natal
1: 250 000
SANBI South African SANBI A series of maps illustrating the type and South Africa
land National Biodi- (Garland et al., 2000) severity of soil degradation between different
degrada- versity Institute land-use types, using qualitative information Magisterial
tion land degradation obtained from 400 extension workers through- districts
review review out SA during 1997 and 1998.

presenting absolute values and are only suitable to prioritise ple, the GLASOD and SANBI studies (shown in Table 1) are
problem areas on a broad scale due to the coarse resolution limited by being lumped for large districts, and due to depend-
(1.1 km) of NOAA images. Another limitation is that both stud- ence on apparently subjective judgments. According to Gibson
ies are based on single-date imagery to test the potential of using (2006; cited in Gibson et al., 2006), the patterns of degradation
remote sensing and GIS as monitoring tools. However, erosion reported in the SANBI study (Garland et al., 2000) are applicable
occurs over a large variety of timescales, such as a single storm only in a relative sense and are difficult to repeat for monitoring
to many decades (Jetten et al., 2003) and single-date imagery purposes. Perhaps the greatest problem with previous regional
does not account for the long-term average soil loss as required assessments of erosion is the lack of comparison and validation
by models such as the USLE. Previous studies not only cover of estimates with actual soil losses.
short or irregular research periods, they also have inconsisten- In order to improve spatial modelling of erosion in SA, a need
cies in their definitions and measurement procedures. For exam- was identified by the DoA to revise model components and tech-

306 Available on website http://www.wrc.org.za

ISSN 0378-4738 = Water SA Vol. 34 No. 3 July 2008
ISSN 1816-7950 = Water SA (on-line)
niques of estimating soil-erosion risk on a national scale. In this et al., 1983; McPhee and Smithen, 1984; Snyman et al., 1986;
context the aim of this study is to improve the spatial soil-erosion Smith et al., 1995; Smith et al., 2000), RUSLE (Haarhoff et al.,
indicators in SA on a national scale, including rainfall erosivity, 1994; Pretorius and Smith, 1998) and the Soil Loss Estimation
soil erodibility, topography and vegetation cover to derive poten- Method of Southern Africa (SLEMSA) developed by Elwell
tial and actual water erosion prediction maps. This study pro- (1976) (Schulze, 1979; Hudson, 1987).
vides a significant update on previous assessments of erosion by Although (R)USLE was originally developed for sub-slope-
inclusion of improved or new national datasets on rainfall, soils, scale soil conservation purposes, the model gained acceptance
topography and vegetation cover which were not available until in regional-scale applications for the following reasons (Lu et
recently. Soil-erosion indicators are further improved by feeding al., 2003):
current available data into advanced algorithms. Each factor is • RUSLE distils soil erosion into a set of measurable primary
assessed as model inputs within a GIS framework and model out- soil-erosion factors that facilitates the input data accessibil-
puts are displayed by means of potential and actual water erosion ity over large regions
prediction maps. Comparison of potential and actual erosion is • The factor-based nature of RUSLE allows easy analysis of
important in policy terms because it indicates those areas which the role of individual factors in contributing to the estimated
are inherently susceptible to erosion (potential risk), but which erosion rate
are presently protected at least to some extent by vegetation • RUSLE has a simple mathematical form facilitating the han-
(actual risk) (Gobin et al., 2003). Results are also validated more dling of large datasets using GIS.
comprehensively than before, followed by a description of the
limitations and challenges that must be overcome in soil-erosion Therefore it was decided to base the current study on a simpli-
assessment on a national scale. fication of RUSLE, the primary function of which is the esti-
mation of (long-term average annual) sheet and rill erosion by
Model selection runoff from slopes in specified cropping and management sys-
tems. The model groups the influences on erosion into five cat-
South Africa covers an area of approximately 121 m. ha and to egories, namely climate, soil profile, relief, vegetation and land
cope with such a large area, analysis must be carried out on a rel- use, and land-management practices; the equation is (Renard et
atively small scale. According to Gobin et al. (2003), the avail- al., 1994):
ability of input data is probably the most important considera-
tion when selecting an erosion model on the regional or national A = R.K.L.S.C.P
scale. It would be impractical to use a sophisticated model if
sufficient input data are not available. On the regional scale, the where:
only means of running a complex model would be to assume A is the spatial average soil loss in t/ha·yr
certain variables and model parameters to be constant (Nearing, R is the rainfall runoff erosivity factor in MJ.mm/ha·h·yr
1998). Prosser et al. (2001) identified this as the dominant rea- K is the soil erodibility factor in t/ha per unit R
son why most soil-erosion prediction carried out on a regional L is the slope length factor
scale is based on empirical relationships. The most well-known S is the steepness factor
and implemented empirical model for estimating soil loss at the C is the cover management factor
regional scale is the Universal Soil Loss Equation (USLE) (Wis- P is the support practice factor
chmeier and Smith, 1978) developed in the 1970s by the United
States Department of Agriculture (USDA), and its upgraded Factor values were estimated from the currently available natu-
version the Revised USLE (RUSLE) (Renard et al., 1994). ral resource data in digital form.
Although developed for application to hill-slopes, the (R)USLE
and its derivatives have been incorporated into many regional- Definitions, methodology and improvements
scale erosion studies across the globe (NRI, 2001; Gobin et al.,
2003; Lu et al., 2003). In South Africa, empirical models have A water erosion prediction map was determined through process-
also been the most widely applied including the USLE (Crosby ing and creating a series of images that represent the RUSLE

Figure 1
Methodology flow
chart for mapping
potential and actual
water erosion

Available on website http://www.wrc.org.za 307

ISSN 0378-4738 = Water SA Vol. 34 No. 3 July 2008
ISSN 1816-7950 = Water SA (on-line)
components in digital form (GIS) (see Fig. 1). The manner in was thus improved by using more detailed digital elevation data
which soil- erosion indicators are classified and improved for (20 m instead of 70 m or higher); and refining the flow trac-
South Africa follows. ing using the combined flow algorithm instead of the single flow
algorithm used before. In addition, the soil and slope factors
Rainfall erosivity (R). The R-factor is the mean annual sum of were separately accounted for, instead of grouping them into
individual storm EI30 values (E is the total storm kinetic energy one, such as in Pretorius (1998).
in MJ/ha·mm and I30 is the maximum 30 min rainfall intensity in A potential water erosion map of SA is generated by combin-
mm/h). However, reliable and long-term information on rainfall ing the above indicators, and represents the inherent susceptibil-
intensity is not available at a regional level and it is necessary to ity of the soil to rainfall erosion, irrespective of vegetation cover
estimate rainfall erosivity from daily rainfall. Here, daily rainfall or land use. Actual soil-erosion risk, which relates to the current
data (Agrometeorology Staff, 1984-2000) was used as input to risk of erosion under present vegetation and land use conditions,
the daily rainfall erosivity model developed by Yu and Rosewell was accounted for as follows:
(1996a and 1996b) in Australia where it was shown to accurately
predict the R-factor and its seasonal distribution. Australia has Vegetation cover index (C) The C-factor is the ratio of soil
a climate that, similar to SA, ranges spatially between winter loss from an area with specified cover and management to soil
rainfall areas in the southwest to a summer rainfall with tropical loss from an identical area in tilled continuous fallow. How-
influences over the northern parts, while large areas over the ever, since it is not possible to take field measurements at a
interior of both countries are classified as semi-arid. Since rain- national scale throughout the year, it was necessary to ascer-
fall is measured at fixed points (weather stations), the inverse tain how crops change with time by means of remote-sensing
distance weight method was used to interpolate data to an EI30 techniques and other sources of literature (e.g. Acocks, 1988;
surface at 2 km resolution for the entire SA. Using more detailed Low and Rebelo, 1998; National Land Cover, 2000). The widely
(stations) and more recent rainfall data than before (e.g. Smithen, used NDVI was used in this study as an indicator of vegetation
1981) an improved rainfall erosivity algorithm was derived that growth determined from images between 2000 and 2004 from
also compensates for topographical influences. the Moderate Resolution Imaging Spectroradiometer (MODIS).
MODIS is more advanced than NOAA data previously used with
Soil erodibility (K) The K-factor may be estimated from data regard to its spatial (250 m 2) and spectral (36 bands) resolution.
on the soil particle size distribution, organic matter content, sur- Subsequently, C-values were assigned through regression equa-
face structure and profile permeability using the soil erodibility tions between vegetation cover and MODIS-derived spectral
nomograph (Wischmeier and Smith, 1978). In the absence of index. The C-factor was estimated using the equations based
soil analytical data in digital form, two alternative sources of on data from Wischmeier and Smith (1978). Assessment of the
soil information were utilised: Soil maps (Soil Survey Staff, support practice factor (P) was excluded by setting the P-fac-
1973-1987) were used to obtain soil erodibility ratings for the tor to 1. Thus, the estimated soil-loss rate for cropping lands
individual soil series of the Binomial Soil Classification Sys- reflects erosion rates with no support practices other than cover
tem of SA (MacVicar et al., 1977); and erodibility values were management. More detail on these procedures is provided by
linked to corresponding soil series in the Land Type Inventories Morgenthal et al. (2006) and Le Roux et al. (2006). Finally, an
(Land Type Survey Staff, 1972-2006) in order to be spatially actual water- erosion prediction map was derived by combining
displayed on a scale of 1:250 000. Using the Soil Loss Estima- C-values with the physical indicators of erosion susceptibility
tor of Southern Africa (SLEMSA) model, soil erodibility units mentioned above.
were assigned based on an assessment of the surface soil texture,
surface soil structure, profile permeability and soil depth of the Results and discussion
dominant soils. Subsequently, the SLEMSA erodibility factors
were used as a guide to the assignment of RUSLE K-factors (in Due to the extensive number of input parameters the RUSLE fac-
SI units t/ha per unit R) to all land types of SA. Previously, this tor maps are provided elsewhere (Le Roux et al., 2006) but the
methodology was only used at a provincial scale or for smaller end product of all the input data and erosion factors is presented
areas, including the Mpumalanga and Gauteng Provinces as well in the accompanying water erosion prediction maps. Two indi-
as ISRDS nodes (e.g. Wessels et al., 2001a; 2001b; Ströhmenger cators are proposed as measures of the area affected by erosion:
et al., 2004). extent to which the total area (e.g. rough estimations per prov-
ince in 106 ha) is affected by water erosion, and percentage of
Topography factors (LS) The effects of topography include the area. Maps are also expressed in quantitative terms and defined
effects of slope steepness (S) and slope length (L). LS-factor into soil-loss classes adopted from Bergsma et al. (1996) in
maps were extracted from 20 m resolution DEMs (GISCOE, t/ha·yr: very low (0 to 5); low (5 to 12); moderate (12 to 25); high
2001) by means of the widely used stream power equation of (25 to -60); very high (60 to -150); and extremely high (>150).
Moore and Burch, (1986; Moore and Wilson, 1992). The main
difference between this equation and the RUSLE LS equation Potential water-erosion prediction map
is the use of upslope contributing area in place of flow-path
length. The stream power equation is the most widely used Partially solving the RUSLE equation using climate erosivity,
method for the extraction of stream networks; to accumulate soil erodibility and topography, provides the erosion suscepti-
the contributing area upslope of each pixel through a network bility or potential soil-erosion risk of the physical environment.
of cell-to-cell drainage paths (Band and Moore, 1995; Gallant Figure 2 thus represents the worst possible situation, which is
and Wilson, 2000). Flow-path lines are constructed from flow the inherent susceptibility of soil to rainfall erosion, irrespective
direction given by an aspect angle. In this study, flow tracing of vegetation cover or land use. The area of land with a moder-
was calculated using a flow algorithm (combined) available in ate to extremely high erosion risk totals approximately 61 m. ha
HydroTools (Schäuble, 2003), which is an add-in program for (50%). Figure 2 clearly illustrates that the eastern parts of the
ArcView GIS 3.x. Methodology from previous erosion studies country has a much higher erosion potential than the western

308 Available on website http://www.wrc.org.za

ISSN 0378-4738 = Water SA Vol. 34 No. 3 July 2008
ISSN 1816-7950 = Water SA (on-line)
 Figure 2
Potential water
erosion-risk
map of South
Africa

part of the country. These areas are mostly associated with hill
and mountain ranges, regions of cyclonic rain and erodible soils.
Conversely, a little over 56 m. ha (46%) of the country is clas-
sified as having a low to very low erosion risk, mainly in the
Northern Cape (29 m. ha; 13.7%) and North-West Province (7 m.
ha; 3.3%) (see Fig. 3). Areas of low erosion risk tend to coincide
with level plateau areas with low rainfall erosivity.

Actual water erosion prediction map

According to the RUSLE, the product of the potential water ero-

sion risk with the cover factor provides the actual water erosion
prediction map of SA (see Fig. 4). As the data in Fig. 3 indicate,
the area of land with an extremely high erosion risk totals over
1 million ha (over 1% of the land surface). Although more than
91 m. ha (75%) are classified as having a very low to low risk,
approximately 26 million ha (20%) of land is eroded at a rate
greater than the suggested soil loss tolerance of 10 t/ha·yr (dis-
cussed under validation). In quantitative terms, the average pre-
dicted soil loss rate for SA is 12.6 t/ha·yr. It should be stressed
that results give a broad overview of the general pattern of the
relative differences, rather than providing accurate absolute ero- Figure 3
sion rates. It is also noteworthy that differences between sedi- Potential and actual erosion risk of each province expressed as
ment yield and soil loss can be very high (Garland et al., 2000). a percentage
Research findings of Scott and Schulze (1991) suggest that soil
loss within a catchment can be up to five times greater than sedi- or more (National Land Cover, 2000). The areas predicted to
ment yield due to the reduction of the total eroded volume by be greatly affected by soil loss when compared to the National
deposition within the catchment. Consequently, a soil-erosion Land Cover appear to be the degraded unimproved grasslands.
figure of 12.6 t/ha·yr could correspond with a sediment yield of Unimproved grasslands are associated with subsistence agricul-
2.5 t/ha·yr. ture where overgrazing of livestock has been excessive. These
Compared to Australia, the average predicted soil-loss rate regions occur widely along the eastern marginal zone, approxi-
for SA is three times as much than that estimated (4.1 t/ha·yr) mately 42 m. ha positioned between the interior plateau and the
by Lu et al. (2003). SA has a higher soil-loss rate than Aus- coast, 0 to 1 200 m a.m.s.l. At the provincial level, the Eastern
tralia presumably due to extensive cultivation and overgrazing. Cape makes the largest (28%) contribution to soil loss. As is evi-
A total of 62% of the country is currently under commercial and dent from Fig. 3, about one third (16 m. ha, 37%) of the province
subsistence farming, including areas that have slopes of 10% is classified as having moderate to extremely high soil loss.

Available on website http://www.wrc.org.za 309

ISSN 0378-4738 = Water SA Vol. 34 No. 3 July 2008
ISSN 1816-7950 = Water SA (on-line)
 Figure 4
Actual water
erosion-risk map
of South Africa

Comparison between potential and actual water et al., 2003). Only recent regional-scale studies are used for gen-
erosion eral comparison (see Table 1), since the geographic coverage of
field- or plot-scale studies is incomplete and cannot provide the
Comparison of the potential risk with the actual soil-erosion risk comprehensive information required of this study. In general,
indicates those areas which are inherently susceptible to erosion, the distribution of actual erosion risk broadly follows that out-
but which are presently protected by vegetation. It is recognised lined previously. Very large percentages of the Eastern Cape,
that there is a huge difference between actual and potential soil Limpopo and KwaZulu-Natal Provinces are under severe threat
erosion, especially along the eastern marginal zone, because low of erosion, whereas Gauteng, the Northern Cape and North-West
C-values (good cover) compensate for the high potential erosion Provinces seem to be the least threatened by water erosion. The
risk. Almost 67% of marginal land has a moderate to severe study by Rooseboom et al. (1992) of sediment yield is worthy of
erosion potential (>12 t/ha·yr), whereas approximately 23% is particular note, as it is based on measurements of fluvial sedi-
classified as having a moderate to severe actual erosion risk. ment loads and covers a wide geographic area. As with findings
Many of these areas are associated with areas of rapid population here, results indicated that some of the highest sediment-yielding
growth and agricultural intensification, and are thus likely to be areas in SA are situated in the north-eastern Cape and southern
at risk. For example, KwaZulu-Natal has large areas of moderate Free State, as well as certain areas of KwaZulu-Natal. It appears
to extremely high potential erosion risk (90%) but relatively low that areas of pronounced relief tend to have the highest soil-loss
actual erosion risk (18%) due to current vegetation cover. The rates, including large tracts of the Drakensberg, the former Tran-
potential erosion map identifies areas of high soil-erosion poten- skei and Waterberg Plateau. This predicted trend is also consist-
tial within some of the natural vegetated areas (e.g. Drakensberg ent with the measurements of Garland et al. (2000) who assessed
area), but these are natural conditions in steep lands experiencing different land-use types at a national scale in terms of the main
high intensity rainfall, and do not produce elevated soil-erosion types of soil degradation affecting them. Rill, and gully erosion
rates. Such comparisons serve to emphasise the importance of are the most important types of land degradation on the commu-
vegetation cover for soil-erosion control, and the dangers inher- nal grazing lands of the eastern parts of the country, especially
ent in changes in land use practice. Over 26 m. ha (at least 30% along the escarpment and coastal plain. The study of Pretorius
of national land) would be subject to high erosion risk with- (1998) also indicates that high soil-loss rates follow the topog-
out maintenance of the current vegetation cover and land use. raphy in certain areas with steep terrain, especially along the
Importantly, around 4.7 m. ha (37%) of cultivated land surface escarpment.
in SA falls in the high to extremely high potential erosion class. The predicted results, however, are not in agreement with
Agricultural intensification could change the land cover, leading all the surveys and areas in SA. Disagreements are evident in
to poorer vegetation cover which is the major pressure indica- areas with grazing and subsistence farming on steep slopes.
tor for soil erosion. The following section compares results with Wessels et al. (2001a; 2001b) and Ströhmenger et al. (2004) pre-
general erosion patterns of erosion risk previously produced. dict high soil-loss rates for these areas in Mpumalanga, Gauteng
and the OR Tambo and Umkhanyakude ISRDS Nodes located
Comparison with previous studies in northern Eastern Cape and KwaZulu-Natal. Current results
indicate that not all subsistence farming areas with steep slopes
Other than visual comparison of maps, there are very few pat- are affected by high erosion rates. Large areas in the OR Tambo
tern comparison techniques available at a regional scale (Jetten node, for example, are not affected by erosion. These regions

310 Available on website http://www.wrc.org.za

ISSN 0378-4738 = Water SA Vol. 34 No. 3 July 2008
ISSN 1816-7950 = Water SA (on-line)
 Table 2 the maximum rate of soil erosion that can occur and still permit
Error matrix between actual erosion map and crop productivity to be sustained economically. McPhee and
observpoints Smithen (1984) proposed a range of soil-loss tolerances in SA
No erosion Row total
between 3 t/ha·yr for shallow soils and 10 t/ha·yr for deep allu-
Erosion
vial soils. In the current study, areas with very low to low soil
n (>10 t/ha·yr) 1 767 1 947 2 714 loss will have calculated erosion rates close to below the highest
n (<10 t/ha·yr) 2 408 7 168 7 576 possible soil-loss tolerance of 10 t/ha·yr. Conversely, areas with
Column total 1 175 9 115 10 290 moderate to extremely high soil loss will have calculated erosion
Omission3 0.65 0.78 rates above the soil-loss tolerance of 10 t/ha·yr. Second, field
Commision4 0.28 0.94 observations mentioned above were separated into points where
erosion was observed and points where no erosion was observed.
Total accuracy 0.77
In achieving this objective, assumptions were made that all ero-
1. Number of points on the actual water erosion prediction map that sion was noted during the surveys and that the current situation
have less than 10 t/ha·yr soil loss
is largely unchanged since these surveys in terms of soil ero-
2. Number of points on the actual water erosion prediction map that
have more than 10 t/ha·yr soil loss sion. Finally, points where erosion was observed were correlated
3. Sample points that have not been correctly classified and have been with areas on the map with moderate to extremely high soil-
omitted from category loss values, whereas points where no erosion was observed were
4. Sample points that have been incorrectly commissioned into correlated with areas on the map with very low to low soil-loss
another category values.
In this context, the error matrix shown in Table 2 indicates
have a high potential erosion risk but a low actual erosion risk that the overall accuracy of the actual water erosion prediction
due to good vegetation cover. Current observations indicate that map is 77%. For points where no erosion was observed, a dis-
erosion sites occur commonly in subsistence farming areas on tinctly higher number of points (7 168) have very low to low
soils with high erodibility values. The results of Rooseboom et erosion compared to points (1 947) where erosion was observed.
al. (1992) support the concept that areas with erodible soils tend For points where erosion was observed, 408 points have very
to yield most suspended sediment. Flügel et al. (2003) confirm low to low erosion compared to 767 points where erosion was
this trend in the Mkomazi catchment in KwaZulu-Natal where observed. Modellers tend to emphasise the successful part of
erosion sites in informal settlements are mainly located on soils the simulation only, while more can be learned from difficul-
with high erodibility values. ties encountered. Therefore, the following section highlights the
More disagreements are evident in arid areas. Pretorius major constraints of the data and lists several factors that should
(1998) predicts much higher erosion rates for the Great Karoo to be taken into account in such a study.
region in the Northern Cape compared to the current study. Pos-
sible explanations include the low rainfall and erosivity values Limitations
for this region, leading to low predicted rates of erosion found
here. Although sheet, rill and gully erosion occur commonly in This study features high levels of spatial and temporal aggre-
large parts of the Karoo, several of these are relict erosion fea- gation and incorporation of a relatively small number of casual
tures. It is postulated that erosion features in some of these areas variables. First, the factors influencing soil erodibility are com-
are of considerable age and may not be contributing to current plex and are influenced by several soil properties. Some of these
sediment yields (e.g. Sneeuberg uplands north of Graaff-Reinet) properties such as organic matter content, stoniness and clay
(Boardman et al., 2003). Other disagreements are noticeable for dispersibility were excluded during estimation of the K-factor in
the savannah region in northern Limpopo and Northern Cape. this study, since the range of descriptive information available
Pretorius (1998) predicts a more severe erosion risk for this region for each soil type is limited at a national scale. Laker (2004)
compared to the current study. His results may be reasonable states that important factors of soil erodibility, such as the parent
since field observations indicate that arid area ground cover is material, degree of soil weathering and stability against disper-
frequently less than its projected vegetation crown cover, which sion and crusting, should not be excluded in modelling. Second,
is not always protective against erosion. C-values for savannah validation of the results indicates that the soil-erosion risk seems
in northern Limpopo and Northern Cape remain questionable to be overestimated for the very steep mountain ranges of the
due to the dense tree canopy concealing the poor ground cover Western Cape and Limpopo Provinces. Although several stud-
when monitored by satellite. Nevertheless, the distribution of ies in SA and across the globe demonstrate that soil erosion is
the actual erosion risk broadly follows that outlined previously. very sensitive to the topographical factor of RUSLE (Biesemans
Such comparisons, however, are not sufficient since the stud- et al., 2000), additional work is still needed to test and validate
ies differ in their definitions and measurement procedures. By the suitability of topography indices in SA and how it affects soil
way of validation, the actual water-erosion map was compared erosion in the country. It appears that the inherent erodibility of
to data collected during field observations (n = 10 290) including the soil and parent material is the overriding erosion-risk factor
the national Land Type Survey (Land Type Survey Staff, 1972 in South Africa, and not the slope gradient, as determined in
to 2006) and verification of the National Land Cover (2000) map the US.
of SA. Another problem of the regional approach followed is the
high variability in space and time of vegetation cover including
Validation data such as ground cover, type of land use, and protection meas-
ures. For example, C-values for Fynbos in the Western Cape are
First, the erosion map was divided into two classes of severity, probably too high, leading to over-estimated soil-erosion values.
but not into different erosion types since the soil-erosion maps do This problem occurs during vegetation senescence when veg-
not distinguish between erosion types. The two severity classes etation indices usually decrease even when the cover remains
are expressed in proportion to typical soil-loss tolerance values; the same (French et al., 2000). However, senescent vegetation

Available on website http://www.wrc.org.za 311

ISSN 0378-4738 = Water SA Vol. 34 No. 3 July 2008
ISSN 1816-7950 = Water SA (on-line)
offers the same protection to the soil as green vegetation and it refinement will be possible given additional research, includ-
is important also to detect relatively dry vegetation. Further- ing:
more, this study calculates mean annual erosion, an approach • The production of more accurate erodibility maps at a
that neglects important seasonal patterns of rainfall erosivity national scale by incorporating key factors such as clay dis-
and cover. More specifically, coincidence of erosive rains with persibility and parent material
low cover in some regions can be a strong control on the mean • Application of RUSLE on a monthly averaged basis by calcu-
annual soil-loss rates. Finally, the RUSLE-based approach will lating appropriate erosivity and cover factors for each month
probably underestimate soil losses in regions where gully and (in order to capture seasonal variations in soil erosion)
subsurface erosion is prominent (Biesemans et al., 2000). These • New high-resolution satellite imagery such as Syste`me Pour
errors, however, can only be challenged at the detailed level (e.g. l’Observation de la Terre (SPOT 5) for detecting individ-
1: 10 000 or small catchment scale). ual erosion features, especially gully erosion from local to
regional scales
Conclusion and recommendations • Establishment of a methodological framework to guide and
standardise future regional soil-loss modelling and mapping
This study based soil-erosion prediction on the principles and efforts. In conclusion, regional studies should combine the
components defined in RUSLE because it combines sufficient simplicity required for application on a regional scale with a
simplicity for application on a national scale with a proper incor- proper incorporation of the most important processes. The
poration of the main soil-erosion factors. It also represents a development of methods that preserve information across
standardised approach and was chosen because of the availabil- scales or quantify the loss of information with changing
ity of spatial input data on each of the soil-erosion factors at a scales has become central in erosion studies.
national scale. Indicators of erosion, including climate erosivity,
soil erodibility, topography and vegetation cover were improved Acknowledgements
over earlier assessments by feeding current available data into
advanced algorithms. Two maps are presented; an actual erosion- The authors would like to thank Mr AJ van Zyl (formerly at
risk distribution, and a potential erosion-risk map that excludes the ARC-ISCW) for initiating the project. The Department of
the vegetation cover factor. Comparison of potential and actual Agriculture, Directorate Land Use and Soil Management (Mr
erosion is important in policy terms because it indicates those HJ Lindemann) supported and funded the research for which we
areas which are inherently susceptible to erosion (potential risk), are thankful. The authors would like to thank Mrs M van der
but which are presently protected by vegetation (actual risk). Walt and Mrs E van den Berg at the ARC-ISCW for providing
Large areas of high potential risk occur in KwaZulu-Natal, the data required for application in the study. Thanks also to
the Eastern Cape and Mpumalanga Provinces, mostly associated Dr HJ Smith, Mr TS Newby, Mr JL Schoeman and Mr HL
with hill and mountain ranges, regions of cyclonic rain and erod- Weepener of the ARC-ISCW who gave valuable advice. The
ible soils. Approximately 50% (61 million ha) of national land help and constant encouragement given by Mr CE Steyn is
has a moderate to severe erosion potential (>12 t/ha·yr), whereas gratefully acknowledged. Editing by Dr TP Fyfield is very much
approximately 20% (26 million ha) of land is classified as hav- appreciated.
ing a moderate to severe actual erosion risk, exceeding the pro-
posed soil-loss tolerance value of 10 t/ha·yr. Comparison of the References
potential and actual erosion risk indicates that over 26 million
ha (30% of national land) would be subject to high erosion risk Acocks JPH (1988) Veld types of South Africa. Memoirs of the
without maintenance of the current vegetation cover. The East- Botanical Survey of South Africa. Government Printer, Pretoria,
ern Cape Province makes the largest (28%) contribution to soil South Africa. 57 1-146.
Agrometeorology Staff (1984-2000) ARC-ISCW Agromete-
loss with approximately one third (16 million ha, 37%) of the
orology Weather Station Network Data for South Africa, Unpub-
province classified as moderate to extremely high. lished. ARC-Institute for Soil, Climate and Water, Pretoria, South
The distribution of the actual erosion risk broadly follows Africa.
that outlined previously; high soil-loss rates follow the topog- Band LE and Moore ID (1995) Scale: Landscape attributes and
raphy in certain areas with steep terrain, especially on the com- geographical information systems. In: Kalma JD and Sivapalan M
munal grazing lands of the eastern parts of the country along the (eds.) Scale Issues in Hydrological Modeling, John Wiley & Sons,
escarpment and coastal plain. Results, however, are not in agree- Chichester, UK. 159-180.
ment with all the previous studies; current results appropriately Bergsma E, Charman P, Gibbons F, Hurni H, Molden-
hauer WC and Panichapong S (1996) Terminology for Soil
indicate that not all subsistence farming areas with steep slopes
Erosion and Conservation. ISSS, ITC, ISRIC.
are affected by high erosion rates. Rather, erosion sites occur Biesemans J, Van Meirvenne M and Gabriels D (2000)
commonly in subsistence farming areas on soils with high erod- Extending the RUSLE with the Monte Carlo error propagation tech-
ibility values. Results are also validated more comprehensively nique to predict long-term average off-site sediment accumulation.
than before, indicating an overall accuracy of 77%. Certain J. Soil Water Conserv. 55 35-42.
obvious anomalies (e.g. Karoo, Fynbos and savannah regions) Boardman J, Parsons AJ, Holland R, Holmes PJ and
reflect the lack of more accurate soil and vegetative cover data Washington R (2003) Development of badlands and gullies in
for SA. This study features high levels of spatial and temporal the Sneeuberg, Great Karoo, South Africa. Catena 50 165-184.
Bühmann C, Beukes DJ and Turner DP (2006) Clay mineral
aggregation and incorporation of a relatively small number of
associations in soils of the Lusikisiki area, Eastern Cape Province,
casual variables. The national-scale information presented here and their agricultural significance. S. Afr. J. Plant & Soil 23 (2)
cannot be used to make decisions at a small-scale (farm-scale or 78-86.
on a pixel by pixel level). Crosby CT, Smithen AA and McPhee P J (1983) Introduction of
Despite these limitations, results remains useful for regional the Universal Soil Loss Equation in the Republic of South Africa,
evaluation and serve as an important basis for the determination Paper No. 83.2072. Proc. of the 1983 Summer Meeting of the ASAE.
of areas where soil conservation should be emphasised. Further 26-29 June, Bozeman, Montana, USA.

312 Available on website http://www.wrc.org.za

ISSN 0378-4738 = Water SA Vol. 34 No. 3 July 2008
ISSN 1816-7950 = Water SA (on-line)
GOBIN A, GOVERS G, JONES R, KIRKBY MJ, KOSMAS C and McPhee PJ and Smithen AA (1984) Application of the USLE in the
GENTILE AR (2003) Assessment and Reporting on Soil Erosion, Republic of South Africa. Agric. Eng. S. Afr. 18 5-13.
Background and Workshop Report, Technical Report 94. European Merritt WS, Letcher RA and Jakeman AJ (2003) A review of
Environment Agency, Copenhagen, Denmark. http://reports.eea. erosion and sediment transport models. Environ. Modell.Software
europa.eu/technical. 18 761-799.
Elwell HA (1976) Natal Agricultural Research Bulletin No. 7, Soil Moore ID and Burch GJ (1986) Physical basis of the length-slope
Loss Estimator for Southern Africa. Department of Agricultural factor in the Universal Soil Loss Equation (USLE). Soil Sci. Soc.
Technical Services, Natal, South Africa. Am. J. 50 1294-1298.
Flügel W, Märker M, Moretti S, Rodolfi G and Sidro- Moore ID and Wilson JP (1992) Length-slope factors for the
chuk A (2003) Integrating geographical information systems, Revised Universal Soil Loss Equation (RUSLE) simplified method
remote sensing, ground truthing and modelling approaches for of estimation. J. Environ. Qual. 22 458-466.
regional erosion classification of semi-arid catchments in South Morgan RPC (1995) soil Erosion and Conservation (2nd edn.). Long-
Africa. Hydrol. Process. 17 929-942. man Group Limited, London, UK.
FRENCH AN, SCHMUGGE TJ and KUSTAS WP (2000) Discrimi- Morgenthal TL, Malherbe J, Van Zyl D and Newby TS
nation of senescent vegetation using thermal emissivity contrast. (2006) Evaluation of MODIS Satellite Data as an Assessment Tool
Remote Sens. Environ. 74 249-254. to Audit the State of Agricultural Resources. ISCW Report No.
Gallant jc and wilson jp (2000) Primary topographic attributes. GW/A/2006/72. ARC-Institute for Soil, Climate and Water, Preto-
In: Wilson JP and Gallant JC (eds.) Terrain Analysis, Principles and ria, South Africa.
Applications. John Wiley & Sons, Inc., New York, USA. 51-85. National Land Cover (2000) Unpublished Beta Release Sup-
Garland GG, Hoffman MT and Todd S (2000) Soil degrada- plied by ARC and CSIR, Pretoria, South Africa.
tion. In: Hoffman MT, Todd S, Ntshona Z and Turner S (eds.) A Nearing MA (1997) A single continuous function for slope steepness
National Review of Land Degradation in South Africa. South Afri- influence on soil loss. Soil Sci. Soc. Am. J. 61 917-919.
can National Biodiversity Institute, Pretoria, South Africa. 69-107. NRI (2001) National Resources Inventory, 2001 Annual NRI, Soil Ero-
http://www.nbi.ac.za/landdeg. sion. United States Department of Agriculture – National Resources
GIBSON DJD (2006) Land Degradation in the Limpopo Province, Conservation Service, USA. http://www.nrcs.usda.gov.
South Africa. M.Sc. Dissertation, University of the Witwatersrand, Oldeman LR, Hakkeling RTA and Sombroek WG (1991)
Johannesburg, South Africa. Glasod World Map of the Status of Human-Induced Soil Degrada-
GIBSON DJD, PATERSON DG, NEWBY TS, LAKER MC and HOFF- tion (2nd revised edn.) ISRIC, Wageningen, UNEP, Nairobi.
MAN MT (2006) Chapter 4: Land. In: Lickindorf E and Clark R Pretorius DJ (1995) The Development of a Soil Degradation Man-
(eds.) South Africa Environment Outlook. A Report on the State of agement Support System. Unpublished ISCW Report. ARC – Insti-
the Environment. Department of Environmental Affairs and Tour- tute for Soil, Climate and Water, Pretoria, South Africa.
ism, Pretoria, South Africa. 87-106. Pretorius DJ (1998) The Development of Land Degradation Moni-
GISCOE (2001) GISCOE Digital Terrain Models. GISCOE, South toring and Auditing Techniques with the Aid of Remote Sensing and
Africa. GIS Technology. ISCW Report No. GW/A/98/27. ARC – Institute
Haarhoff D, Smith HJ, Beytell JF and Schoeman JL (1994) for Soil, Climate and Water, Pretoria, South Africa.
The Testing of Techniques for the Determination of the Erosion Pretorius DJ and Bezuidenhout CJ (1994) Report on the Devel-
Potential of Areas in South Africa. ISCW Report No. GW/A/94/6. opment of a Methodology to Determine the Nature, Rate and Extent
ARC-Institute for Soil, Climate and Water, Pretoria, South Africa. of Soil Erosion in South Africa. ISCW Report No. GW1A/94/7. ARC
HOFFMAN MT and ASHWELL A (2001) Nature Divided. Land – Institute for Soil, Climate and Water, Pretoria, South Africa.
Degra­dation in South Africa. University of Cape Town Press, Cape Pretorius DJ and Smith HJ (1998) The development of a GIS
Town, South Africa. based land information system for developing areas. Proc. of the 21st
HOFFMAN MT and TODD S (2000) A national review of land degra­ SSSSA Conf. 21-23 January, Drakensberg, South Africa.
dation in South Africa: The influence of biophysical and socio-eco- Prosser IP, Rutherford ID, Olley JM, Young WJ, Wall-
nomic factors. J. South. Afr. Stud. 26 (4) 743-758. brink PJ and Moran CJ (2001) Large-scale patterns of erosion
Hudson CA (1987) A Regional Application of the SLEMSA in the and sediment transport in river networks, with examples from Aus-
Cathedral Peak Area of the Drakensberg. M.Sc. Dissertation, Uni- tralia. Mar. Freshwater Res. 52 81-99.
versity of Cape Town, Cape Town, South Africa. Renard KG, Foster GR, Weesies GA, McCool DK and
Jetten V, Govers G and Hessel R (2003) Erosion models: quality Yoder DC (1994) RUSLE Users Guide. Predicting Soil Erosion
of spatial patterns. Hydrol. Process. 17 887-900. by Water: A Guide to Conservation Planning with the Revised Uni-
Laker MC (2004) Advances in soil erosion, soil conservation, land versal Soil Loss Equation. USDA, Agriculture Handbook No. 703,
suitability evaluation and land use planning research in South Washington DC, USA.
Africa. S. Afr. J. Plant & Soil 21 345-368. Rooseboom A, Verster E, Zietsman HL and Lotriet HH
Land Type Survey Staff (1972-2006) Land Types of South (1992) The Development of the New Sediment Yield Map of South
Africa: Digital Map (1:250 000 scale) and Soil Inventory Data- Africa. WRC Report No. 297/2/92. Water Research Commission,
bases. ARC – Institute for Soil, Climate and Water, Pretoria, South Pretoria, South Africa.
Africa. Schäuble H (2003) Hydrotools 1.0 for ArcView 3.x. Technical Uni-
Le Roux JJ, Morgenthal TL, Malherbe J, Smith HJ, versity of Darmstadt, Darmstadt, Germany. http://arcscripts.esri.
Weepener HL and Newby TS (2006) Improving Spa- com.
tial Soil Erosion Indicators in South Africa. ISCW Report No. Schulze RE (1979) Soil loss in the key area of the Drakensberg – A
GW/A/2006/51. ARC – Institute for Soil, Climate, and Water, Pre- regional application of the Soil Loss Estimation Model for Southern
toria, South Africa. Africa (SLEMSA). In: Hydrology and Water Resources of the Dra­
Low AB and Rebelo AG (1998) Vegetation of South Africa, Lesotho kensberg. Natal and Regional Planning Commission, Pietermaritz-
and Swaziland. Department of Environmental Affairs and Tourism, burg, South Africa. pp. 149-167.
Pretoria, South Africa. SCOTT DF and SCHULZE R (1991) The hydrological effects of a
Lu H, Prosser IP, Moran CJ, Gallant JC, Priestley G and wildfire in a eucalypt afforested catchment. S. Afr. Forestry J. 160
Stevenson JG (2003) Predicting sheetwash and rill erosion over 67-74.
the Australian continent. Aust. J. Soil Res. 41 1037-1062. Smith HJ (1999) Application of empirical soil loss models in southern
MACVICAR CN, DE VILLIERS JM, LOXTON RF, LAMBRECHTS Africa: A review. S. Afr. J. Plant & Soil 16 (3) 158-163.
JJN, LE ROUX J, MERRYWEATHER FR, VAN ROOYEN TH, Smith HJ, Haarhof D and Schoeman JL (1995) Die Kwantita-
VON M HARMSE HJ and VERSTER E (1977) Soil Classification: tiewe Beraming van Bronverlies Onder Verskillende Toestande in
A Binomial System for South Africa. Department of Agricultural die Sekhukhune-Distrik in Lebowa. ISCW Report No. GW/W/96/17.
Technical Services, Pretoria, South Africa. ARC – Institute for Soil, Climate and Water, Pretoria, South Africa.

Available on website http://www.wrc.org.za 313

ISSN 0378-4738 = Water SA Vol. 34 No. 3 July 2008
ISSN 1816-7950 = Water SA (on-line)
Smith HJ, Van Zyl AJ, Claasens AS, Schoeman JL and in the OR Tambo and Umkhanyakude ISRDS Nodes. ISCW Report
Laker MC (2000) Soil loss modelling in the Lesotho Highlands No. GW/A/2003/47/2. ARC-Institute for Soil, Climate and Water,
Water Project catchment areas. S. Afr. Geogr. J. 82 64-69. Pretoria, South Africa.
SMITHEN AA (1981) Characteristics of Rainfall Erosivity in South Vrieling A (2006) Satellite remote sensing for water erosion assess-
Africa. M.Sc. Dissertation, University of Natal, Pietermaritzburg, ment: A review. Catena 65 2-18.
South Africa. Wessels KJ, Van den Berg HM, Van der Merwe JPA,
SMITHEN AA and SCHULZE RE (1982) The spatial distribution in Smith HJ, Van Zyl AJ and Twyman L (2001a) Mapping and
Southern Africa of rainfall erosivity for use in the universal soil loss Monitoring the Conservation Status of the Natural Resources of
equation. Water SA 8 (2) 74-78. Mpumalanga Province by Means of Remote Sensing and GIS Tech-
Snyman HA, Van Rensburg WLJ and Opperman DPJ (1986) nology. ISCW Report No. GW/A/2001/9, ARC – Institute for Soil,
Toepassing van ‘n gronderosievergelyking op natuurlike veld van Climate and Water, Pretoria, South Africa.
die Sentrale Oranje-Vrystaat. J. Grassl. Soc. South. Afr. 3 4-9. Wessels KJ, Van der Merwe JPA, Smith HJ, Ströhmenger
Snyman HA (1999) Soil erosion and conservation. In: Tainton NM PHE, Van Zyl AJ and Twyman L (2001b) Gauteng Natural
(ed.) Veld Management in South Africa. University of Natal Press, Resources Audit. ISCW Report No. GW/A/2001/41, ARC – Institute
Scottsville, South Africa. 355-380. for Soil, Climate and Water, Pretoria, South Africa.
SOIL SURVEY STAFF (1973-1987) Soil Profile Descriptions and Soil Wischmeier WH and Smith DD (1978) Predicting Rainfall Ero-
Analyses Data. In: ARC-ISCW Soil Profile Information System, sion Losses, a Guide to Conservation Planning. U.S. Department
Unpublished. ARC – Institute for Soil, Climate and Water, Pretoria, of Agriculture, Agricultural Handbook No. 537, Washington DC,
South Africa. USA.
Ströhmenger PHE, Van der Merwe JPA, Smith HJ, Van Yu B and Rosewell CJ (1996a) An assessment of a daily rainfall
den Berg HM, Van den Berg EC, Dekker C, Malherbe erosivity model for New South Wales. Aust. J. Soil Res. 34 139-152.
J, De Nysschen G, Van der Walt M, Haasbroek P, Yu B and Rosewell CJ (1996b) Rainfall erosivity estimations using
Morgenthal TL, Kellner K and Van Rensburg L daily rainfall amounts for South Australia. Aust. J. Soil Res. 34
(2004) Auditing the Conservation Status of the Natural Resources 721-733.

314 Available on website http://www.wrc.org.za

ISSN 0378-4738 = Water SA Vol. 34 No. 3 July 2008
ISSN 1816-7950 = Water SA (on-line)