Impact of Overgrazing on Diffuse and Concentrated Erosion: Case Study in the Sloping Lands of South Africa

: Soil erosion is one of the most critical threats to cultivated land. Yet little information is available in Sub-Saharan Africa, especially on the relative contributions of various forms of erosion. Therefore, this study’s objective was to quantify soil loss by sheet and linear erosion. The study was carried out on the sloping land rangeland of the Potshini catchment of KwaZulu-Natal, South Africa, with an annual average rainfall of 766 mm. The average sheet erosion computed using a network of 1 m 2 microplots was 7.7 ton ha − 1 y − 1 with standard error of 1.97 ton ha − 1 y − 1 (which corresponded to an ablation rate of between 0.35 to 1.32 mm y − 1 ) while linear erosion, mainly the retreat of gully banks, removed 4.8 ton ha − 1 y − 1 , i.e., 38.4% of total soil losses. Despite removing a lower amount of soil, sheet erosion by depleting fertile, carbon- and nutrient-enriched soil horizons has a great impact on most ecological functions associated with soils.


Introduction
Erosion by water is one of the main causes of soil loss and impacts food production [1,2]. Soil erosion by water occurs as sheet and linear erosion. Sheet erosion removes a thin layer of soil from the soil, which has long been unnoticed, but since the early 1990s, it has been identified as one of the major threats to the sustainability of soils [3]. Linear erosion takes place when runoff concentrates and can lead to the formation of rills and gullies.
Most often, sheet and linear erosion have been investigated independently [4], and several questions have arisen: (1) What is the relative contribution of splash, sheet and gully erosion to the overall soil erosion? (2) What are the mechanisms involved and controlling factors? (3) What are the interactions between the different types of water erosion? (4) What are the appropriate remediation techniques at landscape level?
In South Africa, as in many areas of the world, the evaluation and prediction of soil erosion has often been conducted using the universal soil loss equation (USLE) method [5]. However, this method, even in its revised version, does not account for the contribution of linear erosion as pointed out by [6], thus leading to incorrect estimations. When based on field investigations, fluxes of sediments are evaluated within rivers such as by [7] reporting soil erosion rates of 306 g m −2 y −1 at 23 ha level and 0.29 g m −2 y −1 at 1000 ha level, but with no information on the respective contributions of sheet and linear erosion.
Linear erosion and especially deep gullies, locally termed 'dongas', are a main concern in South Africa because of their number [17][18][19][20][21] and because they pose severe constraints on the agricultural use of land by rendering much of the farmland inaccessible. Dongas mainly affect steep slope areas dedicated to rangeland [22]. Despite there being numerous studies on gully erosion that have provided information on gully occurrence (e.g., [21]), there is limited quantitative information on the amount of erosion involved. Dongas are associated with the presence of certain parent materials [19], such as unconsolidated colluvions associated with high sodium contents [23,24], and erode easily [20]. Ref [25], using rainfall simulation to assess the mechanisms of gully bank retreat, showed that the dominant mechanism is runoff detachment followed by splash and the fall down of aggregates.
In this context, the main objective of this work is to compare sheet and linear erosion at the catchment level. This study was conducted within an agricultural area under communal grazing and sloping lands of South Africa and specifically at the Potshini catchment, where several research studies have been performed on linear erosion and several others on sheet erosion. Here, we compared the findings by [15], which were obtained using 1 × 1 m runoff plot, and those by [26] on linear erosion using a network of benchmarks at gully banks and gully head cuts.

Geology, Climate and Soils
The area is a 1000 ha catchment (longitude: 29.36 • ; latitude: 28.82 • ) located within the KwaZulu-Natal province, South Africa ( Figure 1). It is under rangeland (highlands) and croplands lowland), and the climate is sub-tropical humid whose particularity is that most of rains occur in summer [27]. The mean annual precipitation from the nearest station (Bergville, which is located 10 km to the east of the study site) is 684 mm y −1 , and the potential evaporation is 1600 mm y −1 . The mean annual temperature is 13 • C [27]. Altitudes at the study site ranges from 1080 to 1455 m, and the topography is relatively gentle with a mean slope gradient of 16%. Soils were formed from the Karoo Supergroup, which exhibits a succession of fine-grained mudstones, siltstones, sandstones and shales [28]. While the rangeland is freely grazed the whole year, lowlands are manually ploughed for rain-fed beans and maize production, and little fertilization and herbicides are used due to the limited access to funds.

Micro-Plots Used for the Estimation of Sheet Erosion
Sheet erosion was assessed using fifteen 1 m 2 runoff plots [15] installed at different topographical positions, soil type and overgrazing density. As can be seen in Figure 1  The soils are classified as Inanda soil (Soil Classification Working Group, 1991) or Acrisols [29]. Deep Acrisols (>1 m) with a pH of 4.5 and a C content of 1.5% in the 0-10 cm range. A horizon characterizes bottomlands and footslopes, while slopes are characterized by shallow soils (<0.5 m, with a pH of 4.5 and a C content of 1.2%) because of erosion. In the latter, a humiferous A horizon of dark reddish brown (5YR 3/3) color overlays in most bases a massive clayey (40%) Bw horizon (from 30 to 60 cm). In some situations, and due to accelerated erosion, the BW horizon directly outcrops the soil surface.

Micro-Plots Used for the Estimation of Sheet Erosion
Sheet erosion was assessed using fifteen 1 m 2 runoff plots [15] installed at different topographical positions, soil type and overgrazing density. As can be seen in Figure 1, three plot replicates were used per modality: "Deep Acrisols at footslope (F); shallow Acrisols at midslope (M); Shallow Acrisols at terrace (T); Deep reddish Acrisols at shoulder under dolerite (SD); and deep yellowish Acrisols at shoulder under sandstone (SS)" (From [15]).
The metal borders surrounding the plots were inserted in the soil to 0.1 m, and the field measurements were carried out over two years. These were thus assumed to occur under steady-state conditions, as no significant soil cracking or linear erosion occurred. After each runoff event, the total runoff volume (R) was collected and evaluated using graduated containers, and a 800 mL aliquot was sampled for the estimation of sediment concentration (SC) through drying at 110 • C for 48 h. Finally, the total soil losses (SL) were computed as the product of R by SC.
A total of 555 samples were collected during the study, during which 37 runoff events occurred. Each rainstorm was characterized by the amount of rain, the maximum and average rainfall intensity (over 6 min).

Linear Erosion
Gully bank and gully head cut retreats had a major impact at the study site. Both the survey of gully bank and gully head cut retreats were performed using erosion pins. Erosion pins consist of metallic nails that are set into the gully banks so that the top of the pin gives a datum from which changes in the length that protrudes above the soil surface can be recorded for assessing rates of soil loss over time ( Figure 2). This method was adapted from a widely used method developed for soil surface erosion by [30] and is readily available. Measurements of the length that protrudes above the soil surface can be made in a very flexible way, from the storm event level to every 10 years. The 0.15 m pins were inserted into the soil to the soil surface to set a benchmark for soil erosion. Their diameter was of 5 mm, and a metallic washer was placed at their top to ensure first finding the nails after each rain event and second to ensure a more accurate reading of the gully bank retreat. Secondary 0.5 m long and 10 mm thick nails ( Figure 2) were also inserted into gully banks to register large erosion rates, such as these due to the collapse of gully banks.
Pins were installed along soil profiles distant by 10 m and for each profile, they were inserted horizontally at respectively 0.15, 0.5 and 1 m from the soil surface. The 0.5 m long metallic bar was installed at 0.5 m deep (Figure 2). Hundred and ten soil profiles were equipped at 3 gullies, totaling 1076 m of gully banks and 440 data points. The pins were installed in September 2007 and surveyed in April 2008 to estimate the average gully bank retreat to be converted into an average yearly soil loss.
banks to register large erosion rates, such as these due to the collapse of gully banks.
Pins were installed along soil profiles distant by 10 m and for each profile, they were inserted horizontally at respectively 0.15, 0.5 and 1 m from the soil surface. The 0.5 m long metallic bar was installed at 0.5 m deep (Figure 2). Hundred and ten soil profiles were equipped at 3 gullies, totaling 1076 m of gully banks and 440 data points. The pins were installed in September 2007 and surveyed in April 2008 to estimate the average gully bank retreat to be converted into an average yearly soil loss.

The Rainfall Conditions
The 2007-2008 rainy season was characterized by 37 rainfall events causing erosion. The cumulative rainfall amount of these 37 events was 766 mm, and the yearly rainfall amount was 850 mm. The average rainfall amount (RA) of these events was 21 mm with values between 4 and 49 mm, and the average 6 min rainfall intensity (RI) was 13 mm h −1 (4 < RI < 38 mm h −1 ). The larger events occurred on the 11th of December (RA = 48 mm; RI = 9 mm h −1 ), 14th of January (RA = 43 mm; RI = 14 mm h −1 ) and 21st of April (RA = 48 mm; RI = 22 mm h −1 ). Twenty-five events had a RA of more than 15 mm and a RI of more than 15 mm h −1 , which might be considered highly erosive events.

Soil Inter-rill Erosion
The average runoff coefficient (R) computed from the 37 erosive rainfall events using the 15 microplots was 25% with values between 0 and 98%. The median R was 19%, and fifty percent of data were between 8 and 37%. The average sediment concentration was 3 g L −1 with a median at 1 g L −1 , and 50% of the data were between 0 and 3 g L −1 . The most extreme value was 51 g L −1 and occurred on 21 st April 2008. The resulting average soil

The Rainfall Conditions
The 2007-2008 rainy season was characterized by 37 rainfall events causing erosion. The cumulative rainfall amount of these 37 events was 766 mm, and the yearly rainfall amount was 850 mm. The average rainfall amount (RA) of these events was 21 mm with values between 4 and 49 mm, and the average 6 min rainfall intensity (RI) was 13 mm h −1 (4 < RI < 38 mm h −1 ). The larger events occurred on the 11th of December (RA = 48 mm; RI = 9 mm h −1 ), 14th of January (RA = 43 mm; RI = 14 mm h −1 ) and 21st of April (RA = 48 mm; RI = 22 mm h −1 ). Twenty-five events had a RA of more than 15 mm and a RI of more than 15 mm h −1 , which might be considered highly erosive events.

Soil Inter-Rill Erosion
The average runoff coefficient (R) computed from the 37 erosive rainfall events using the 15 microplots was 25% with values between 0 and 98%. The median R was 19%, and fifty percent of data were between 8 and 37%. The average sediment concentration was 3 g L −1 with a median at 1 g L −1 , and 50% of the data were between 0 and 3 g L −1 . The most extreme value was 51 g L −1 and occurred on 21 April 2008. The resulting average soil losses from the 37 events was 2.9 g. The first quartile was at 1 g and the third at 17 g, and the maximum value was 1247 g (Figure 3).
The average yearly soil loss by sheet erosion was 7.7 ton ha −1 y −1 . This was much higher than the country average of 3 ton ha −1 y −1 estimated by Whyte (1995). However, the sheet erosion at our site was much lower than estimations made under similar sloping land conditions in Asia under similar Acrisols, where losses were between 6 and 35 ton ha −1 y −1 [31,32]. At our site, soil losses by sheet erosion ranged between 3.5 ton ha −1 y −1 at shoulders to 13.2 ton ha −1 y −1 at the terrace. The footslope position lost 5.3 ton ha −1 y −1 , which was the second lowest erosion rate ( Table 1). The shoulder, with 9.9 ton ha −1 y −1 , was the second most erosive position. The greater erosion rate by sheet erosion at the middle slope correlated with the highest sediment concentration (5.6 g L −1 versus and average of 4.1 g L −1 ) and high runoff (30 vs. 23%). On the contrary, lower soil erosion rates correlated with lower runoff and sediment concentration as observed at the shoulder. It was, however, interesting to note that as the erosion rate at the terrace was close to the rangeland average, its runoff was much lower (16.5 vs. 23%) while its sediment concentration was much higher (5.0 vs. 4.1 g L −1 ). Such trends might be explained by the presence of well structure soils with the predominance of stable clayey aggregates that are easily removable and transportable by water erosion.
Hydrology 2022, 9, x FOR PEER REVIEW 5 of 9 losses from the 37 events was 2.9 g. The first quartile was at 1 g and the third at 17 g, and the maximum value was 1247 g (Figure 3). The average yearly soil loss by sheet erosion was 7.7 ton ha −1 y −1 . This was much higher than the country average of 3 ton ha −1 y −1 estimated by Whyte (1995). However, the sheet erosion at our site was much lower than estimations made under similar sloping land conditions in Asia under similar Acrisols, where losses were between 6 and 35 ton ha −1 y −1 [31,32]. At our site, soil losses by sheet erosion ranged between 3.5 ton ha −1 y −1 at shoulders to 13.2 ton ha −1 y −1 at the terrace. The footslope position lost 5.3 ton ha −1 y −1 , which was the second lowest erosion rate ( Table 1). The shoulder, with 9.9 ton ha −1 y −1 , was the second most erosive position. The greater erosion rate by sheet erosion at the middle slope correlated with the highest sediment concentration (5.6 g L −1 versus and average of 4.1 g L −1 ) and high runoff (30 vs. 23%). On the contrary, lower soil erosion rates correlated with lower runoff and sediment concentration as observed at the shoulder. It was, however, interesting to note that as the erosion rate at the terrace was close to the rangeland average, its runoff was much lower (16.5 vs. 23%) while its sediment concentration was much higher (5.0 vs. 4.1 g L −1 ). Such trends might be explained by the presence of well structure soils with the predominance of stable clayey aggregates that are easily removable and transportable by water erosion.   Linear Erosion Soil losses (ton ha −1 y −1 ) 4.8

Soil Linear Erosion
The average gully bank retreat (both laterally and upstream) estimated from the 440 pins installed vertically at the 110 gully bank profiles selected was 0.037 m, i.e., 3.791 cm y −1 . The median was at 1.5 cm, and 50% of the values were between 0.5 and 3.0 cm. The differences between the pins installed at 0.15, 0.5, and 1 m deep were not significant. Thirty-six pins showed a retreat of more than 15 cm reaching in two situations about 100 cm, which corresponded to the collapse of entire gully banks. A much higher occurrence of gully bank collapse occurred in the downslope parts. The landscape at the place of deep Acrisols resulting in an average retreat of 6.7 cm, compared to 2.0 cm for the middleslope and 1.9 cm for the upslope.
The total gully bank length was 1076 m, and the average gully bank height was 3.0 m. Considering that an upstream drainage area of these gullies is 23 ha and an average soil bulk density of 1.3 Mg m 3 , the estimated yearly erosion rate by linear erosion was 4.8 ton ha −1 . This is a relatively low rate compared with the existing literature. [33] in their review of the work done worldwide on linear soil erosion reported erosion rates as high as 65 ton ha −1 y −1 such as in the badlands areas of Spain. Studies in Africa reported rates of 32 ton ha −1 y −1 in Niger and 15 ton ha −1 y −1 in the neighboring country of Lesotho [33], which were much higher than this observed in the present study. The erosion rate of 4.8 ton ha −1 y −1 was, however, of the same order as that found under tropical conditions in Kenya (3 ton ha −1 y −1 , [34] and much higher than the 1.3 ton ha −1 y −1 lost from the sloping lands of Laos [35]. A possible reason for such a low level of linear erosion is the high stability of soil aggregates (which was previously observed by [15] as compared to the studies investigating for instance low aggregate stability soils, such as those found in the loess of Northern Europe.

Conclusions
This study made a direct assessment of soil losses by sheet and linear erosion within a typical sloping rangeland of KwaZulu Natal in South Africa. Over the 2007-2008 rainy season with a rainfall amount of 766 mm, the overall soil erosion was 12.5 ton ha −1 y −1 . Linear erosion contributed 7.7 ton ha −1 y −1 , corresponding to 61.6% of total soil losses, while sheet erosion averaged 38.4% or 7.7 ton ha −1 y −1 (Figure 4). These results on the predominance of linear erosion confirm previous observations made in tropical areas, where this erosion form is considered dominant in the process of soil degradation and downstream damages, such as dam siltation [36].
The second conclusion is that sheet erosion, which was highly variable within the 23 ha area (from 3.5 to 13.2 ton ha −1 y −1 ), corresponded to a soil ablation rate of 0.35 to 1.32 mm y −1 that would be 4.5 to 366 times higher than the natural rock weathering and soil formation [37,38]. From a qualitative point of view, the removed soil material is the most fertile, carbon-and nutrient-rich part of the soil with direct consequences on soil ecological functions of food and biomass production, water storage, support of biodiversity and/or climate change.
Overall, gullies represent a surface area of 6411 m 2 (i.e., about 2.8% of the whole rangeland surface area). Considering an average gully bank retreat of 3.79 cm as found in this study, the loss of agricultural land increases each year by 28 m 2 (i.e., 0.4%). This might not seem at first sight a staggering amount, but extrapolated to the next 10 years and assuming a similar erosion rate, the loss of agricultural land might be as much as 4%. Moreover, while sediments eroded in the slopes by sheet erosion are likely to resettle within the slopes [39,40], the soil material detached from the gully banks is in direct contact with the stream flow and is thus likely to constitute the major part of the exported sediments from the area. Up to 85% of sediment loads to a reservoir in a 0.73 km 2 catchment in the tropical regions of southern China have been shown to have a gully bank origin [41]. The proportion of the exported sediments from rangelands originated from sheet and linear erosion needs to be further investigated. More is to be done as well on the temporal variations of sheet and linear erosion and on the quality of the exported sediments and their impact on downstream ecosystems. The second conclusion is that sheet erosion, which was highly variable within the 23 ha area (from 3.5 to 13.2 ton ha −1 y −1 ), corresponded to a soil ablation rate of 0.35 to 1.32 mm y −1 that would be 4.5 to 366 times higher than the natural rock weathering and soil formation [37,38]. From a qualitative point of view, the removed soil material is the most fertile, carbon-and nutrient-rich part of the soil with direct consequences on soil ecological functions of food and biomass production, water storage, support of biodiversity and/or climate change.
Overall, gullies represent a surface area of 6411 m 2 (i.e., about 2.8% of the whole rangeland surface area). Considering an average gully bank retreat of 3.79 cm as found in this study, the loss of agricultural land increases each year by 28 m 2 (i.e., 0.4%). This might not seem at first sight a staggering amount, but extrapolated to the next 10 years and assuming a similar erosion rate, the loss of agricultural land might be as much as 4%. Moreover, while sediments eroded in the slopes by sheet erosion are likely to resettle within the slopes [39,40], the soil material detached from the gully banks is in direct contact with the stream flow and is thus likely to constitute the major part of the exported sediments from the area. Up to 85% of sediment loads to a reservoir in a 0.73 km 2 catchment in the tropical regions of southern China have been shown to have a gully bank origin [41]. The proportion of the exported sediments from rangelands originated from sheet and linear erosion needs to be further investigated. More is to be done as well on the temporal variations of sheet and linear erosion and on the quality of the exported sediments and their impact on downstream ecosystems.