Organic Carbon and Aggregate Stability of Three Contrasting Soils as Affected by Arable Agriculture and Improved Pasture in Northern KwaZulu-Natal, South Africa

Changes in soil organic carbon (SOC) and aggregate stability as a result of agricultural practices have been extensively studied, but the extent of these changes is site and crop specific. In this study, SOC and aggregate stability were evaluated under undisturbed grassland, cultivated pasture and arable land uses in northern KwaZulu-Natal, South Africa. The soils in this study were sampled from the 0–5, 5–10 and 10–20 cm depths of a Nitisol, Stagnosol and Fluvisol, and analysed for SOC content and stocks and aggregate stability. In all soils, SOC was significantly lower under arable cultivation (range from 1.4 to 2.1%) compared to the grassland and pasture (3.4 to 4.2%) in the top 10 cm. The soil carbon stocks followed a similar trend to the SOC in all soils and under all land uses. The overall aggregate stability expressed as mean weight diameter (MWD) was also significantly lower under the arable land use (1.03 to 1.82 mm) compared to grassland (2.44 to 2.96 mm) at all depths in all the soils, except at 5–10 cm in the Stagnosol. The MWD values under grassland and pasture were equal at all depths in all the soils. The MWD increased with SOC up to a threshold of 3.0–3.9% indicating the importance of SOC in soil aggregation. The weak correlation in the Stagnosol suggests that other factors are important in this soil. The loss of SOC and decline in aggregate stability over a period of 38 years of continuous arable cultivation in the near surface layers of two of the soils are potentially problematic for soil sustainability in the longer term.


Introduction
Native grasslands have been transformed to both cultivated pastures and arable farming by many South African farmers due to the need for new land for cultivation (du Preez and Claassens 1999). The changes in land use create major challenges including increased soil disturbance and a reduced amount of litter added to the soil that lower soil organic matter (SOM), which is a critical determinant of soil biological, chemical and physical properties (Blanco-Canqui and Lal 2004;Mills and Fey 2003). Native grasslands are able to accumulate large amounts of SOM and their conversion to arable farming often results in a decline in soil organic carbon (SOC) or its derivative, the soil carbon stock (SC stock ).
In general, SOC losses from former undisturbed grasslands caused by arable farming practices are estimated to range from 20 to 50% of the initial content within the first 40-50 years of cultivation (du Preez and Claassens 1999; Guo and Gifford 2002;Blanco-Canqui and Lal 2004;Larre-Larrouy et al. 2004). Du  reported a 10 to 75% decrease in SOC in 27 South African soils (0-20 cm) with 1 to 85 years cropping history. Lobe et al. (2011) also recorded a 37% reduction in SOC in the surface layer (0-20 cm) 98 years after native grassland was converted to arable farming in the Free State, South Africa. Similarly, Gajić et al. (2013) reported a 52% decrease in SOC (from 2.27 to 1.75%) of the surface layer (0-30 cm) 10 years after conversion of native grassland to arable cropping in a Luvisol in central Serbia. This was more rapid than the decrease reported by Lobe et al. (2011), possibly due to climatic differences. These losses are generally rapid in the first 20 years and gradually stabilise at a new steady state over the next 50 years (Baldock and Skjemstad 1999;du Preez et al. 2011). Conversely, several studies have noted a substantial increase in SOC after native grasslands were converted to permanent pasture in a wide range of soils (Miles and Hardy 1999;Dominy and Haynes 2002;Milne and Haynes 2004;Boley et al. 2009). According to Haynes and Williams (1993), the increase is a result of an appreciable increase in SOM addition, particularly as litter material, but also as cattle dung (Blanco-Canqui and Lal 2004;Larre-Larrouy et al. 2004). The rate of decline of SOC after conversion of grassland to arable land, and increase after conversion to pasture, could depend on soil type. The effects of land use change from grassland on SOC could also affect soil aggregate stability.
Aggregate stability and SOC have been found to be closely related such that the loss of SOC following virgin soil disturbance impacts negatively on the stability of aggregates (Tisdall and Oades 1982;Six et al. 2000a;Carter 2002). There are, however, contradictory findings about this relationship as some studies have shown that SOC increases aggregate stability (Carter 1992;Chenu et al. 2000), others present evidence to the contrary (Gerzabek et al. 1995;Haynes 2000), while others have reported no relationship (Perfect and Kay 1990;Carter et al. 1994). These conflicting results are most likely due to the type of soil and type of vegetation/crops involved. A reduction in aggregate stability, as measured by the mean weight diameter (MWD), after long-term cultivation has been reported in a number of studies (Haynes and Beare 1997;Carter 2002;Gajić et al. 2013). However, the effects of arable agriculture on aggregate stability have been found to vary with soil type, crop type and management. For example, Shepherd et al. (2001) reported that mica-rich, fine-textured soils showed the greatest decrease in the MWD of aggregates, while oxide-rich soils, and particularly allophanic soils, showed only a slight decrease in MWD after 18-40 years of cropping. In another study, Reid and Goss (1981) found that root growth of perennial ryegrass (Lolium perenne) and lucerne (Medicago sativa) generally increased aggregate stability, while growth of maize (Zea mays), tomato (Solanum lycopersicum) and wheat (Triticum aestivum) decreased it. Dufey et al. (1986) also found that ryegrass increased aggregate stability, while red clover (Trifolium pratense) had no significant effect.
In KwaZulu-Natal Province of South Africa, several studies (e.g. Meyer et al. 1996;Van Antwerpen and Meyer 1996;Qongqo and Van Antwerpen 2000;Dominy et al. 2001;Dominy and Haynes 2002;Graham et al. 2002a;Graham et al. 2002b) have been conducted to quantify the effects of sugarcane (Saccharum officinarum) mono-cropping after more than 30 years since the conversion of the land use from undisturbed grassland. All these studies have reported a significant decrease in SOC and aggregate stability among other factors. While these studies have documented important SOC information concerning the soil degradation as a result of land use change, much remains to be established about the conversion of native grasslands to other land use types such as pasture and vegetable farming, as different crop and grass species affect SOC and soil structure differently (Haynes and Naidu 1998;Gajić et al. 2013). This study therefore compared the effects of improved pasture and arable agriculture with that of undisturbed grassland after 38 years of continuous cultivation on SOC and aggregate stability in northern KwaZulu-Natal.

Site Description
The study was carried out at the Owen Sitole College of Agriculture farm (OSCA) (28° 57′ 45″S; 31° 55′ 31″E) located at Kwesaka-Mthethwa about 12.5 km north of Empangeni and 163 km north of Durban, KwaZulu-Natal Province, South Africa (Fig. 1). The 672 ha farm ranges in altitude from 23 to 120 m a.s.l. and has a subtropical climate with hot, humid summers and cooler, drier winters. The annual daily mean temperature range is 19 to 33 °C with 867 mm annual rainfall (Van der Linden 2004). The farm is dominated by Shortlands (Sd), Westleigh (We) and Inhoek (Ik) soil forms (Soil Classification Working Group 1991, 2018 which are broadly equivalent to Nitisol, Stagnosol and Fluvisol soil groups (IUSS Working Group WRB 2014), respectively. Basalt of the Letaba Formation of the Lebombo Group dominates the underlying geology of the studied area (Van der Linden 2004).

Site History
Based on personal interviews and interpretation of aerial photographs, the predominant land uses in the area were livestock farming and subsistence agriculture between 1937 and 1968. The major portion of the farm was ploughed and intensively grazed and the veld was burned frequently. After the smallholder farmers were moved off the land in 1968, the area was fenced and it reverted to the original natural grassland. This undisturbed grassland had not been cultivated for 38 years prior to this study but the area is patchy. The planted pasture is grazed by cattle and goats and is a productive mixture of kikuyu (Pennisetum clandestinum), Italian ryegrass (Lolium multiflorum) and weeping love grass (Eragrostis curvula). Typical annual fertiliser rates for the pasture are 75 kg N; 15 kg P and 5 kg K ha −1 and the plots are irrigated by a sprinkler system. Since 1968, the arable land has been used mostly for vegetables including cabbage (Brassica oleracea) and beetroot (Beta vulgaris) and field crops such as maize, potatoes (Solanum tuberosum) and dry beans (Phaseolus vulgaris). The land had received annual applications of limestone ammonium nitrate, monoammonium phosphate, potassium chloride and calcitic lime at recommended rates for the crops grown.

Site Selection, Soil Sampling and Preparation
In the absence of previous information for the study area, changes in soil properties induced by land use dynamics were evaluated by taking soil samples from plots of land under pasture and arable land use systems and comparing their physicochemical properties with soils under natural or semi-natural grassland that was less disturbed. A prerequisite for this method was that the reference undisturbed grassland and the land use treatments were located such that differences in geology, topography and climate did not introduce bias into the data. Given this condition, any differences in soil properties are then able to be attributed to the differences in land use.
The emphasis in this study was to ensure that land use types at each site were on similar soil forms hence a preliminary soil survey was carried out. The sampled soils were located in typical landscape positions with the red, structured Sd on upper to middle slopes, the plinthic We on middle slopes, and the black, structured Ik soil form on lower to middle slopes. The Sd and Ik soil forms were deep (> 1 m) while the We was shallow (< 0.4 m). Sampling was carried out from three land uses on the Sd, We and Ik soil forms (Fig. 1). Four replicate soil samples were collected at three depths (0-5, 5-10 and 10-20 cm) from mini-pits (90 cm wide, 1200 cm long and 30 cm deep) in May 2012. Most of each sample from the three sampled depths was crushed to pass a 2-mm mesh. Sub-samples were sieved and the 2.8-5 mm diameter aggregates were collected and air dried prior to measurement of aggregate stability. Undisturbed soil samples for bulk density measurement were collected using stainless steel core rings of 5 cm height and Fig. 1 The location of the sampling points within each land use at Owen Sitole College of Agriculture (OSCA) in KwaZulu-Natal Province, South Africa 7.8 cm in diameter (238.9 cm 3 inner volume) from 0-10 and 10-20 cm depths.

pH, Exchangeable Acidity and Particle Size Distribution
Soil pH was measured in a 1:2.5 soil: 1 M KCl suspension using a standard glass electrode with a Metrohm E396B pH metre. The suspension was stirred and allowed to stand for 30 min before the pH was measured. Exchangeable acidity was measured by titration with 0.005 M NaOH following extraction with 1 M KCl solution (Manson and Roberts 2000). Particle size of the < 2 mm fraction was determined by the pipette method (Gee and Bauder 1986). Sand content was determined using a dry sieving technique and the silt and clay fractions were determined after dispersion and sedimentation (Day 1965). The textural class was determined from the texture triangle (Soil Classification Working Group 1991).

Bulk Density, Organic Carbon and Soil Carbon Stocks
Bulk density was determined by the core method of Blake and Hartge (1986) and organic carbon by the dichromate oxidation method (Walkley 1947). The carbon stocks of the soils were estimated using the proportion of SOC, bulk density and depth increment (Eq. 1).
where SC stock represents the soil carbon stocks (kg C m −2 ), SOC is the soil organic carbon (g C kg −1 soil), BD is bulk density (kg m −3 ) and D is the soil sampling depth (m).

Soil Structural Stability
Soil structural stability was measured by fast wetting 10 g of the 2.8 to 5 mm aggregates to mimic explosive slaking under rapid flooding such as during intense rainfall events. The procedure followed the first step of the French AFNOR norm NF X 31-515 (AFNOR 2005) developed by Le Bissonnais (1996). The aggregates were rapidly added directly into water and allowed to stand for 10 min before the water was extracted by pipetting. The sample was passed through a 0.05-mm sieve and aggregates > 0.05 mm were collected and transferred onto a 50-mm sieve previously immersed in ethanol, and shaken five times with a gentle, regular, helical rotation movement. The > 0.5 mm aggregates on the sieve were collected and dried at 40 °C and then gently dry sieved using a nest of six sieves: 2000, 1000, 500, 200, 106 and 50 µm. The aggregate stability (1) SC stock = SOC × BD × D is represented by the mean weight diameter (MWD) corresponding to the sum of the mass fraction remaining on each sieve (Eq. 2) (Kemper and Rosenau 1986).
where d is the mean diameter between two sieves (mm) and m is the weight fraction of aggregates remaining on the sieve (%).

Statistical Analysis
Statistical analysis was carried using GENSTAT version 18 (Payne et al. 2011). Generally two-way analysis of variance (ANOVA) was executed to compare land uses and soil types on each of the variables measured for each soil. Differences between the means of the significant factor were assessed with Tukey's HSD post hoc test (p ≤ 0.05). The undisturbed grassland, pasture and arable land uses were compared within each soil form using LSD at p ≤ 0.05. The mean values for SOC were plotted against the MWD of aggregates to determine the relationships of these parameters using the dataset of all soils combined and separately for the Sd, We and Ik soils. Pearson's correlation coefficient was used to explore the significant relationships for the soil quality indicators measured and all results were based on four replications in the field.

General Description and Basic Properties of the Soils
The Sd had a clay texture at all depths and under all land uses, although the amount of clay was lower under arable than the other two land uses (Table 1). The We was a sandy clay loam under the undisturbed grassland in the 0-10 cm depth but increased in clay at 10-20 cm to a sandy clay probably due to the restricted drainage caused by the plinthic subsoil. Under the pasture and arable land uses the We generally had a finer texture (clay or clay loam). The Ik was a sandy clay under undisturbed grassland but had a clay texture at all depths under the other two land uses (Table 1). All the soils at all depths and under all land uses were acidic with pH ranging from 4.83 to 5.56 with the only sample outside this range being the 10-20 cm depth of the Ik under undisturbed grassland with a pH of 6.39 (Table 1). Exchangeable acidity was very low in all the soils regardless of depth or land use. (

Soil Organic Carbon, Bulk Density and Carbon Stocks
The overall (0-20 cm) SOC mean values under the undisturbed and pasture land uses on the Sd were higher (3.1%) than arable (1.4%), corresponding to a 55% decrease under arable relative to the undisturbed grassland (p < 0.05; Fig. 2a-c). This trend was the same for all depths in the Sd soil. In the 0-5 cm depth of the We soil, the SOC was not significantly different between undisturbed (3.5%) and pasture (3.9%) and greater than the arable (2.2%) land. In the 5-10 cm depth, pasture (4.5%) was greater than the undisturbed (3.0%) and arable (2.0%) while no differences were observed between the land uses in the 10-20 cm depth. The overall SOC mean values in the We soil were 3.3% (undisturbed), 3.6% (pasture) and 1.8% (arable) corresponding to an increase of 9% under pasture and a decrease of 46% under arable relative to the undisturbed grassland (p < 0.05; Fig. 2a-c). As in the Sd, the overall SOC in the Ik soil was higher in the undisturbed (3.9%) and pasture (3.5%) than the arable land (2.0%), corresponding to decreases of 10 and 49%, respectively, relative to the undisturbed grassland (p < 0.05; Fig. 2a-c). Also, similar to Sd, the trend of lower SOC under arable than pasture and grassland was the same at all depths of the Ik soil.
In the 0-20 cm depth of the Ik soil, the BD was not different between undisturbed (0.91 g cm −3 ) and pasture (0.94 g cm −3 ) and both land uses were lower than the arable (1.17 g cm −3 ), corresponding to increases of 29 and 25%, respectively, relative to the undisturbed grassland (p < 0.05; Table 1 pH, exchangeable acidity and particle size distribution of Shortlands, Westleigh and Inhoek soils at 0-5, 5-10 and 10-20 cm depth under grassland, pasture and arable land use (n = 4) Means within a column for the same soil depth followed by different letters are significantly different from others (p ≤ 0.05). Means within a column for the same soil depth not followed by letters are not significantly different (p > 0.05). SCL, sandy clay loam; SC, sandy clay; CL, clay loam  Fig. 3a). The BD was different between the land uses on the Ik for 0-10 cm and 10-20 cm soil depth layers (p < 0.05; Fig. 3b-c) while no differences were observed for Sd and We.
The overall (0-20 cm) SC stock of the Sd followed a similar trend to SOC with undisturbed (5.35 kg C m −2 ) and pasture soils (5.78 kg C m −2 ) higher than those under arable (3.00 kg C m −2 ), corresponding to increases of 44 and 48%, respectively (p < 0.05; Fig. 4a). A similar trend was found for the 0-10 cm depth of the Sd soil (p < 0.05; Fig. 4b). In the 0-20 cm depth of the We, the SC stock was not different between land uses but the surface (0-10 cm) layer had a higher stock under the undisturbed (3.41 kg C m −2 ) and pasture land uses (4.22 kg C m −2 ) than under arable (2.45 kg C m −2 ) (p < 0.05; Fig. 4b). As in the Sd, the overall SC stock in the Ik soil was higher in the undisturbed (6.91 kg C m −2 ) and pasture (6.32 kg C m −2 ) than the arable land (4.12 kg C m −2 ), corresponding to decreases of 8.5 and 40%, respectively, relative to the undisturbed grassland (p < 0.05; Fig. 4a). Also, similar to Sd, the trend of lower SC stock under arable than pasture and grassland was found for the 0-10 cm depth of the Ik soil (p < 0.05; Fig. 4b) while the SC stock was not different between land uses and soil types in the 10-20 cm depth increment (p > 0.05; Fig. 4c).  Fig. 3 The comparison of soil bulk density (BD) (mean ± standard error; n = 4) in Shortlands (Sd), Westleigh (We) and Inhoek (Ik) soil forms with depth under grassland (G), pasture (P) and arable (A) land use systems. Means associated with the same letter are not significantly different (LSD 5% )

Soil Aggregate Stability
The MWD of the three soils under the different land uses are shown in Fig. 5. Overall, all the soils under grassland and pasture land uses had similar aggregate stabilities which were higher than for arable at each depth, except in the We at 10-20 cm. Aggregate stability in all the arable soils was similar at each depth except the 10-20 cm depth where the We was higher than the other two soils. In the Sd, the overall (0-20 cm) MWD value was significantly lower (p < 0.05) under the arable land use (1.03 mm) compared to the undisturbed grassland (2.96 mm) and pasture (2.44 mm). Similarly, in the We the overall MWD values were significantly higher for undisturbed (2.79 mm) and pasture (2.90 mm) land uses than arable (1.82 mm). In the Ik, the overall values were not different between the undisturbed grassland (2.47 mm) and pasture (2.48 mm) but significantly higher than under arable (1.27 mm) (p < 0.05). The land use and soil interaction was significant in the We but not significant in either the Sd or Ik.

Relationship Between Soil Organic Carbon and Aggregate Stability with Selected Soil Properties
The linear relationships of SOC and MWD were poor and only appeared to hold up to about 3.0% SOC for the combined dataset and for the Sd alone, and to about 3.5% for the Ik, above which aggregate stability did not change. The relationships were better described by quadratic functions for the combined dataset of all the soils (Fig. 6a), and the Sd (Fig. 6b) and Ik (Fig. 6d), while that for the We had a poor relationship (Fig. 6c). The quadratic functions showed that the maximum MWD was 2.79 mm (at 3.75% C) for all soils combined, 2.83 mm (at 3.32% C) for Sd and 2.61 mm (at 3.90% C) for Ik. Table 2 gives the overall (0-20 cm) correlation coefficients for selected properties of the three soils under the three land uses. The combined dataset revealed low correlations between all the measured properties except for MWD and SC stock (r = 0.56), MWD and BD (r = − 0.42), SOC and SC stock (r = 0.49) and the interdependence of sand and clay. The SOC positively correlated with SC stock (r = 0.78), MWD (r = 0.69) and silt (r = 0.78) in the Sd, and SC stock (r = 0.45) together with MWD (r = 0.53) in the Ik (Table 3). In these soils, the SOC was also correlated positively with sand (r = 0.45) and negatively with clay (r = − 0.46) only in the Ik. The MWD was correlated with silt (r = 0.65) positively in the Sd and negatively (r = − 0.39) in the We, but not in the Ik. The correlation of MWD with clay was negative (r = − 0.43) in the Ik, and not significant for the Sd and We soils. The BD correlated negatively with MWD in the Sd (r = − 0.49) and Ik (r = − 0.55) while the correlation between MWD and SOC (r = − 0.76) was also negative in the Ik.

Discussion
Continuous cultivation of the same land often results in a decline in pH due to the export of base cations in harvested produce, leaving acidic cations, and oxidation of NH 4 + to NO 3 − and organic residues. However, this effect was not observed in this study. As nitrate is leached, Ca 2+ , Mg 2+ and K + usually serve as counter-ions (Haynes and Francis 1990;Haynes and Swift 1990;Dominy et al. 2001). This leaves a higher concentration of H + in the near-surface layers and the pH falls. In accord with these findings, Covaleda et al. (2006) showed a significant reduction of pH after conversion of the native ecosystem to farmlands in Nitisols in Mexico.
The non-significant differences in pH correlated with no significant differences in exchangeable acidity across all the land use types, indicating that the soils are well buffered. The low exchangeable acidity in the topsoils might be indicating the presence of stable organo-mineral complexes of aluminium with SOM, minimising the possibility of hydrolysis and release of H + ions (Haynes and Swift 1990;Leinweber et al. 1993;Nsbiamana et al. 2004). The significant increase in pH with depth under the grassland is presumably related to accumulation of base cations from the parent material (Graham et al. 1995;Adesodun et al. 2007).
Clay and silt did not differ significantly between land uses in individual soils, which is inconsistent with the results reported by Gebrelibanos and Assen (2013) in the northern highlands of Ethiopia under Luvisols. In their study, higher sand contents and lower clay fractions were recorded after the native vegetation was converted to arable agriculture and this was attributed to the selective removal of clay particles by processes of erosion leaving behind the sand fraction. The differences in the results may perhaps be reflecting differences in environmental conditions of the study areas, i.e. the greater elevation (1800-2500 m a.s.l.) and precipitation (987 mm) of the northern highlands of Ethiopia could have resulted in higher erosion rates in relation to the elevation (23-120 m a.s.l.) and precipitation (867 mm) of the study area.
The SOC generally decreased from undisturbed grassland to arable land use, with no differences between grassland and pasture across all three soil types, with the exception of the We which had greater SOC under pasture than under grassland. The decline in SOC was probably a result of increased oxidation of SOM exposed from aggregates through tillage operations, coupled with limited organic matter input in the arable lands (Leinweber et al. 1993;Six et al. 2000b;Milne and Haynes 2004;Puget and Lal 2005;Tayel et al. 2010). Many researchers have explained in detail that cultivation breaks soil aggregates and exposes previously inaccessible organic matter to microbial attack, accelerating the decomposition and mineralisation of SOC (Haynes and Naidu 1998;Haynes 1999;Larre-Larrouy et al. 2004;Nsbiamana et al. 2004). Gale et al. (2000) concluded that wider plant spacing and removal of the harvested crop and crop residues result in lower biomass input to arable soils. Dominy and Haynes (2002) also reported an 18% SOC loss from the 0-40 cm layer of an Oxisol profile after native grasslands were converted to maize fields in the midlands of KZN. Compared to the native ecosystem, Freixo et al. (2002) found a 62% decrease in SOC content (from 45 to 17 g C kg −1 ) in the surface layer (0-20 cm) of a Rhodic Ferralsol after conversion to arable in southern Brazil. Similarly, Boajilla and Gallali (2010) sampled 13 different soils (0-20 cm) under different climates in Tunisia and found that SOC content was 83% lower in cultivated soils compared to native grasslands. The combination of low SOC content together with higher BD contributed to the lower SC stock observed in arable soils (Blanco-Canqui and Lal 2004;Bronick and Lal 2005). The SC stock results obtained in this study are also comparable to Guo and Gifford (2002) who performed a meta-analysis of 74 studies from a wide range of soils under native ecosystems and cultivated lands and found that cultivation decreased the SC stock by 42%. The decline in SOC is considered to reduce aggregate stability and thus increase the erodibility of the soil.
The similar concentration of SOC under pasture compared to native grassland probably reflects the minimal disturbance causing an immediate increase in SOC (Six et al. 1998;Miles and Hardy 1999;Boley et al. 2009). Under these land use systems, soil disturbance by cultivation is reduced and very large inputs of organic material, particularly as above-ground litter, but also as root turnover is added to the soil (Mills and Fey 2003;Boley et al. 2009;Gebrelibanos and Assen 2013). This, together with increased microbial activity and a lower surface soil temperature commonly associated with less disturbed environments, has been observed to improve SOC content while reducing the BD compared to arable management systems (Six et al. 2000a(Six et al. , 2000bBronick and Lal 2005;Puget and Lal 2005). These results are, however, contrary to those of Boajilla and Gallali (2010) who observed a higher SOC content under native grassland than pasture in Tunisia. The difference could again be related to differences in climatic conditions. When compared to the warmer, wetter conditions of northern KZN, the hot, dry Mediterranean summers and mild winters of Tunisia favour less biomass production.
Similar to the SOC, the MWDs were lower in the soils under arable than the other two land uses because of decline in aggregate stability associated with lower SOC. The MWDs measured are of similar magnitude to those found by Boajilla and Gallali (2010) who reported 2.79, 2.10 and 1.70 mm from the uppermost soil layer (0-20 cm) under virgin grassland, pasture and a mixed crop management system, respectively. Emadi et al. (2009) also found a 67% decrease in the MWD of water-stable aggregates for a soil depth of 0-30 cm over 20 years of active cropping practices in Typic Haploxerolls of Iran (Mediterranean climate). The results are, however, higher than those of Gajić et al. (2013), who found MWDs of 0.91 mm under native grassland compared to 0.81 mm under arable agriculture in Luvisols of central Serbia. The differences in the results of the two studies are probably due to the different soil types and crop species at the studied sites. Luvisols generally have weak structure compared to the soils in the present study. Broersma et al. (1997), studying the effects of diverse cropping systems on aggregation in a Luvisol in Canada, reported that crops affect soil structure differently because of diverse rooting habits, rhizosphere processes, amount and type of additions and their ability to provide surface soil protection. Similar findings were reported by Haynes and Beare (1997) after studying the influence of six crop species on aggregate stability in the Canterbury region of New Zealand.
In the present study, the lower MWD values of the arable soils may be due to reduction in the stover-derived organic binding agents required for the formation of stable aggregates upon crop residue removal (Blanco-Canqui and Lal 2004;Lal 2009). On the other hand, regular irrigation of the pasture favoured the accumulation of SOM to an almost similar level to that of the undisturbed grassland (Leinweber et al. 1993;Haynes and Naidu 1998;Six et al. 2000a). Planted grass species enhance rainwater infiltration and favour a more even microclimate beneath their canopy. These conditions should generate a more active fauna and flora and, as a consequence, higher aggregate stability (Le Bissonnais 1996;Bronick and Lal 2005;Lal 2009). The lowest average MWD value (0.78 mm) was recorded in the 0-5 cm layer of the We under arable indicating that this soil is generally unstable with very frequent risk of surface sealing, overland flow and interrill erosion (Le Bissonnais et al. 2002). In contrast, the average MWD values greater than 1.8 mm shown by the undisturbed and pasture land uses indicate stable to very stable soils (Le Bissonnais et al. 2002).
The poor relationships observed between SOC and clay in all the studied soils, including a negative relationship in the Ik, were not expected as clay acts as a binding agent, aggregating particles together and so influences SOC decomposition through physical and chemical adsorption processes (Hassink 1997;Amezketa 1999;Bronick and Lal 2005). These results may nevertheless be reflecting the inability of these soils to form organo-mineral complexes with strong bonds (Blanco-Canqui and Lal 2004;Adesodun et al. 2007;Lal 2009).
The breakdown of a linear relationship between SOC and MWD above about 3.0% C for all soils and the Shortlands alone, and above 3.5% C for the Inhoek, suggests that any SOC higher than these levels will not result in a corresponding increase in MWD. This indicates that organic C becomes less important for aggregate stability at higher levels. These results reflect earlier findings by Tisdall and Oades (1982) who reported that above a certain SOC concentration there is no further increase in the aggregation effect, although no specific threshold value was given. The SOC and MWD values found in this study are very similar to those reported by Mthimkhulu (2011)  . This study has shown that the positive relationship of SOC and aggregate stability is significant up to a maximum MWD of 2.60-2.80 mm at SOC contents ranging from 3.0 to 3.9% for the studied soils in the locality.
The positive correlation between MWD and SOC found in the Ik (r = 0.78) and Sd (r = 0.69) soils is similar to relationships observed in other studies elsewhere (Idowu 2003: r = 0.76;Milne and Haynes 2004: r = 0.68; Boajilla and Gallali 2010: r = 0.72), and is considered to reflect the central role of organic matter in the formation and stabilisation of soil aggregates (Hartemink 1997;Six et al. 2000b;Bronick and Lal 2005). The poor relationship between SOC and MWD for the We could be explained by the relatively high aggregate stability close to or above 2.27 mm, which was the maximum for all the soils and thus was not responsive to SOC. The weak relationship may also be a result of the middle slope position of the We that has resulted in erosion of the upper soil layers thereby bringing less humusrich subsoil nearer to the surface. In accordance with the We results in the present study, Levy and Mamedov (2002) and Six et al. (2004) found weak correlations between SOC and MWD, while Carter (2002), Mthimkhulu (2011) and Madikizela (2014) reported no correlation between SOC and MWD suggesting that other soil properties are also important in the stability of soil aggregates (Amezketa 1999). Unlike the results of Tayel et al. (2010) that showed an increase in the MWD with clay content, the stability of aggregates appeared to be less dependent on clay than SOC and silt content as indicated by stronger correlations between these soil properties ( Table 3). The negative correlation of MWD and silt suggests that higher silt results in weaker aggregates (lower MWD). On the other hand, the SOC and sand content seemed to be the driving factors of aggregate stability in the Ik. In agreement with these results, Idowu (2003) reported a significant linear correlation between MWD and silt (r = 0.69) and fine sand (r = 0.65) under different cultivation practices in Alfisols in Nigeria. According to Norhayati and Verloo (1984) and Igwe et al. (1995) a certain amount of fine sand and silt are needed with clay for the formation of stable aggregates. Sand and silt are, however, not as active as clay in stabilising aggregates owing to their lower surface area and charge density (Lado and Ben-Hur 2004). The negative relationship observed between MWD and BD in the Sd (r = − 0.49) and Ik (r = − 0.55) soils is also common and attributable to enhancement of aggregation by SOC thereby resulting in a reduction in the BD (Graham et al. 1995;Haynes 1999Haynes , 2000Milne and Haynes 2004). In the Ik soil, SOC was also negatively correlated with BD (r = − 0.76).
Although the generally weak correlations observed in this study cannot be definitely related to clay mineralogy, there is evidence from other reported work that the relationship between clay content and aggregate stability is not always linear as the clay type is also vital in aggregation (Baldock and Skjemstad 1999;Nciizah and Wakindiki 2014). For example, Lado and Ben-Hur (2004) observed lower MWD (0.25 mm) in soils dominated by phyllosilicate minerals even though they had higher clay (63%) content compared to soils with about 30% clay that were dominated by non-phyllosilicate minerals (MWD = 0.84 mm) from different locations in Israel and Kenya. Nciizah and Wakindiki (2014) also found poor correlations between MWD and clay content in topsoils (0-20 cm) from 14 ecotopes of a semi-arid to subhumid region in the Eastern Cape Province, South Africa. It was concluded that the effect of soil mineralogy on aggregate stability is dependent on the morphology or structure of each mineral.

Conclusions
Long-term cultivation of vegetables and field crops reduced aggregate stability, organic carbon and carbon stocks in arable soils than under pasture and undisturbed grassland suggesting increased susceptibility to slaking and soil erosion. The pastureland improved or maintained aggregate stability and soil organic carbon content when compared to undisturbed grassland. Soil organic carbon had a positive correlation with aggregate stability but not with other soil properties indicating that most soil functions are not affected by only the organic carbon content. The lack of correlation between mean weight diameter and clay content across all the studied soils suggested that the clay mineralogy might be more important in affecting aggregate stability. Thus, an understanding of the dominant clay minerals in each soil would provide a more rigorous basis for predicting soil behaviour in the context of land use change, at least in general terms.