The study was carried out in the Nkuhlu large-scale, long-term exclosures situated in the southeastern region of the Kruger National Park of South Africa (Figure 1), approximately 18 km from Skukuza (24°59′10″ S, 31°46′24.6″ E). The area receives approximately 550 mm of precipitation on an annual basis and the average temperature is 21.9 °C, with temperatures dropping and rising to an average of 5.7 °C and 32.6 °C during the winter and summer seasons, respectively. The exclosures were established in 2002 on the premise that by virtue of their size, large and tall mammalian herbivores (mainly elephants and giraffes) are the most significant drivers of change and maintenance in environments they inhabit (O’Keefe & Alard 2002; Scogings 2014). They cover roughly 139 ha of land that has been sub-divided into three separate units, namely: (1) full exclosure, a 70 ha fully fenced site where animal entry is prohibited, split into a burnt and unburnt plot (2) partial exclosure, a 44 ha partially fenced site where large and tall animals, particularly elephants and giraffes are not permitted, split into a burnt and unburnt plot (3)open access area (control site), a 25 ha unfenced area where all animals are allowed (Jonsson et al. 2010). No burning takes place in the control site.

The current study was conducted in the unburnt plots of full exclosure and control site. Nkuhlu exclosures are situated in sloping land that extends from the crest to the Sabie River channel, which is situated on the foot-slope and/or valley bottom (Janecke 2020; O’Keefe & Alard 2002). Soils within the crest and foot-slope vary between shallow sandy soil directly overlying hard granite rock and deep sandy to sandy loam Na-dominated soil, respectively (Khomo & Rogers 2005; Scogings et al. 2014). The World Reference Base (WRB) classifies the shallow sandy soil on the crest as Leptosol, while the South African system (Soil Classification Working Group [SCWG]2018) classifies them as Mispah and shallow Glenrosa. The deep, Na-rich duplex soil on the foot-slope can be classified as Luvisol in the WRB or Oakleaf and Montagu soils according to the South African Soil Classification system (Siebert & Eckhardt 2008). This paper will only consider the crest. According to Scogings, Hjältén and Skarpe (2011), the most common woody species throughout the Kruger National Park, including the crest of Nkuhlu exclosures include: (1) C. apiculatum, (2) Combretum hereroense, (3) G. bicolor, (4) Grewia flavescens, (5) Dichrostachys cinerea and (6) Euclea divinorum. This article only considers the C. apiculatum and G. bicolor species (Figure 2). These two species were selected on the basis that they are among the most abundant in the study area and the fact that their composition and vegetation structure is highly influenced by herbivory (Bakker et al. 2016). Most mammalian herbivores, particularly elephants and giraffes include substantial quantities of C. apiculatum and G. bicolor in their diets by ingesting large amounts as they find them palatable (Seloana et al. 2018). Also, elephants alter the composition and vegetative structure that influences soil properties by congregating under trees, stripping them off and discarding the leaves on the savanna floors.

For each of these two tree species, soil samples were collected at three radial distances from the tree trunk namely: (1) Zone A, tree centre (beside tree trunk) – below canopy, besides the trunk of the tree; (2) Zone B, canopy edge – along the edge of the tree crown and (3) Zone C, outside canopy – open grassland zone or savanna patches in the inter-canopy space.

The experiment was a 3 × 2 × 2 factorial design with different combinations of two herbivore utilisations levels (open access and full exclosure), two woody species (C. apiculatum and G. bicolor species) and three sampling zones (under the canopy, edge of canopy and far from canopy). Five soil samples were collected from three points within each sampling site. As a result, each sampling area had three composite samples (replicates), each comprising five sub-samples randomly collected at a sampling depth of 10 cm. Therefore, a total of 36 samples were collected from Kruger National Park. Subsequently, the collected soil samples were air-dried, mechanically crushed and pulverised manually using a 2-mm sieve. Soils sampled for the purpose of microbial analysis were collected a sampling depth of 5 cm and immediately transferred into sampling bags and stored at approximately 4 °C in a cooler box filled with ice before passing them through a < 2-mm sieve. These soil samples were immediately analysed as Wallenius et al. (2010) reported that even short delays may impact the results.

Dry combustion method using an automatic highly sensitive CN analyser was used to measure TN and TC (Bremner & Mulvaney 1982). Determination of pH of soil samples was carried out using a glass electrode pH meter at a soil-to-water ratio 1:2.5 (w/v) by mixing 5 g of air-dried soil with 25 mL deionised water (McLean 1982). The molybdenum blue method of Olsen and Sommers (1982) was used for measuring available P, and extraction was carried out using sodium bicarbonate (NaHCO₃). Ammonium acetate (NH₄Oac) was used for extracting exchangeable cations (Ca²⁺, Mg²⁺, K⁺ and Na⁺), which were quantified using the inductively coupled plasma spectrometer (ICP-MS). The values for these elements were used to calculate the cation exchange capacity (CEC). The fluorescein diacetate (FDA) hydrolysis method was used to measure total microbial activity in soil as described by Schnürer and Rosswall (1982).

Statistical analyses were completed with JMP statistical software (version 16.0, SAS Institute Inc.). Analysis of variance (ANOVA) was used to analyse treatment effects by assessing whether TC, TN, soil pH, exchangeable Ca, Mg, K and Na, CEC, available P and total microbial activity differed among the three treatments (exclosure, woody species and canopy cover based radial distance). Fischer’s protected least significant difference was used for mean separation at an alpha level of 0.05.

Ethical clearance to conduct this study was obtained from the University of the Free State Environment and Biosafety Research Ethics Committee (no. UFS-ESD2021/0192/21]).

The results of the study revealed that there were statistically significant differences in TN and TC (p < 0.05; Figure 3a and Figure 4a) among the different exclosures and sampling zones, with the open access area and samples under tree canopy having higher TN and TC than the other treatments. Samples collected under tree canopies had significantly higher TN and TC than those collected along the edge of the tree crown and in open grassland zones (Figure 3a and Figure 4a). However, there were no significant differences in TN and TC between samples collected on the edge as well as away from canopy. The two species did not independently cause statistical changes in total N. The two-way interaction between exclosure and sampling zone had a significant effect on TN and TC (Figure 3b and Figure 4b). For G. bicolor in the open access site, it was observed that samples collected under canopy had high TN; however, there were no significant differences between the samples under canopy and those along the edge of the tree crown, as well as between the canopy edge and the open grassland zone. For both woody species, it was also observed that the samples collected under tree canopies in the control site (i.e. with higher herbivore biomass) had the higher TC than all the other treatments (Figure 4c).

Exclosure, woody species and sampling zone did not interactively cause statistical changes in TN for all C. apiculatum treatments and G. bicolor in the full exclosure, but samples outside canopy of G. bicolor in the full exclosure had significantly lower TN than the canopy edge of G. bicolor and the canopy of C. apiculatum in the open access area (Figure 3c). Herbivory, tree species and canopy-based radial distance did not interactively cause statistically significant differences in TC in the full exclosure, but samples collected under canopy of both species in the control site had higher TC than those sampled in the full exclosure (Figure 4c).

Soil pH of all the sampling areas showed that they are slightly acidic. However, exclosure and sampling zone had a significant main effect on soil pH (p < 0.05; Table 1). The full exclosure resulted in significantly higher pH than the control site where herbivores were permitted. Areas under tree canopies had significantly lower pH than tree crown edges and open grassland zones. There were no significant differences in pH between samples collected on the edge of canopy and those collected in the open grassland zone. Woody species also did not have a significant effect on soil pH.

Data regarding exchangeable cations revealed that there were statistical differences in exchangeable K and Mg between the open access area and full exclosure (Table 1). The open access had higher exchangeable K and lower exchangeable Mg than the full exclosure. Exchangeable Na and Ca were not statistically influenced by exclosure. Woody species did not have a significant effect on exchangeable cations (Ca²⁺, Mg²⁺, Na⁺ and K⁺). Sampling zone caused significant changes on exchangeable K and Ca but not Na and Mg, with samples collected under tree canopy having higher exchangeable K and Ca than both the area on the edge and outside of canopy. There were, however, no differences in exchangeable K and Ca between samples on the edge of canopy and those sampled far away from the crown. Cation exchange capacity was significantly affected by exclosure and woody species but not the sampling zone, with the open access area having higher CEC than the full exclosure. Furthermore, G. bicolor had significantly lower CEC than C. apiculatum. We also found that the two-way interaction between exclosure and woody species had a significant influence on exchangeable Mg and CEC but not exchangeable Ca, Na and K. Additionally, exchangeable K and Ca were significantly influenced by the two-way interaction between exclosure and sampling zone but not on other exchangeable cations and CEC. The interaction between woody species and sampling zone as well as the three-way interaction between exclosure, woody species and sampling zone did not cause any statistical changes in exchangeable cations and CEC.

Exclosure had a significant main effect on available P (p < 0.05; Figure 5a), with the open access area having higher available P than full exclosure. All the samples collected besides the tree trunk had high available P relative to other sampling zones. Species did not independently cause statistical changes in available P, nor were there any significant interactions between species and exclosure, as well as between species and sampling zone.

The two-way interaction between exclosure and sampling zone caused statistical changes in available P, with areas under C. apiculatum and G. bicolor in the open access area having much higher available P than all the other treatments (Figure 5b and c). In the full exclosure, it was observed that the interaction between exclosure, woody species and sampling zone did not have any influence on availability. In the open access area, the canopy edge of G. bicolor had significantly higher available P than the area outside of canopy, however, there were no statistical differences in available P between the canopy edge and area under canopy as well as between the area under canopy and the open grassland zone.

The results of the study revealed significant main and interactive effects of exclosure and species on total microbial activity (p < 0.05; Figure 6a, Figure 6b and Figure 6c). The control site (i.e. with higher herbivore biomass) had significantly higher microbial activity compared to the full exclosure (i.e. herbivores not permitted). Sampling zone did not cause any significant changes in microbial activity. However, G. bicolor had significantly higher microbial activity than C. apiculatum (Figure 6a). The interaction between exclosure and woody species caused statistical changes in total microbial activity. The canopy edge of G. bicolor in the open access area had higher microbial activity than all the sampling zones of both C. apiculatum and G. bicolor in the full exclosure.

It was also noticed that all the samples collected under and on the edge of G. bicolor crown in the open access area had higher microbial activity than the area outside of canopy, with areas under canopy being not significantly different from canopy edge and the open grassland zone not being significantly different from the area under canopy (Figure 6c). Samples collected in outside the canopy of the C. apiculatum tree had the lowest microbial activity; however, it was not significantly different from other treatments, except for samples collected under C. apiculatum in both the full exclosure and open access area as well as all the sampling zones of G. bicolor in the control site.

Our experimental study revealed that, with the exceptions of exchangeable Ca and Na, herbivory resulted in prominent changes in all the biochemical properties under investigation, indicative of the crucial role played by animals in maintaining savanna ecosystem function. The exclusion of herbivores caused significant increases in soil pH and exchangeable Mg, while the presence of herbivores led to significant increases in TN and TC, available P, microbial activity, exchangeable K and CEC. Our results coincide with the findings of Abdulahi et al. (2006) who reported higher TN, organic matter and soil P status in herbivore grazing sites of an African savanna in Eastern Ethiopia. The increase in TN and C, available P, microbial activity, CEC and exchangeable K could be attributed to the continuous deposition of animal excreta (Malongweni & Van Tol 2022). In this regard, in the full exclosure where the entry of herbivores is prohibited, soil improvement is a result of nutrient returns from litterfall. However, in the open access area (i.e. where herbivores are permitted), there will be higher nutrient returns because of defoliation, debranching of trees by browsing herbivores and the deposition of animal excrement into the soil (Abdulahi et al. 2016; Holdo & Mack 2014; Mordelet, Abbadie & Menaut 1993).

The results of the study also indicated that herbivore exclusion increased soil pH and exchangeable Mg. These results correspond with the findings of Abdulahi et al. (2016) who reported significantly higher values of soil pH and Mg in grazing sites of a semi-arid Ethiopian savanna. Abdulahi et al. (2016) claim that the increase was likely because of higher levels of CEC. Cation exchange capacity of the soil generally increases with soil pH because of a higher proportion of hydroxide (OH^–) ions present in the soil (Brady & Weil 2022). Although a high soil pH is likely to cause an increase in CEC, the results of our study indicated an inverse relationship between pH and CEC, with the full exclosure having a significantly lower pH and high CEC than the open access area. Nonetheless, higher CEC in the full exclosure was accompanied by a high concentration of exchangeable Mg. This may be caused by a decrease in the number of hydrogen (H⁺) ions caused by soil pH increases.

The main influence of the two tree species did not result in any changes in most of the soil properties under investigation in our study. Significant changes were only observed in soil microbial activity and CEC, with G. bicolor having a significantly higher microbial activity compared to C. apiculatum. This could be because in Kruger National Park and many other semi-arid African savannas, G. bicolor is considered one of the most important forages during the dry season (Chira & Kinyamario 2009; Le Houerou 1980; Siebert & Eckhardt 2008). Mammalian herbivores tend to congregate around it as they find it more palatable than C. apiculatum (Baumer 1983; Brink 2007; Le Houerou 1980). During grazing, large herds of plant eaters deposit significant amounts of urine and dung under the G. bicolor canopy. Dung and urine harbour constituents that promote microbial activity (Holdo & Mack 2014). According to Chira and Kinyamario (2009), G. bicolor is particularly enjoyed by elephants. Elephants tend to break down G. bicolor tree branches when foraging. But this species is known to resprout rapidly. New leaves are often of better quality than the previously browsed leaves because the N and P content of new leaves is often higher. Also, the number of leaves increases after defoliation. Improved forage quality and a higher number of leaves promote microbial activity because of better and higher nutrient returns (Schroder 2021).

Despite having a high microbial activity, G. bicolor had significantly lower CEC than C. apiculatum. These results do contradict the finding of Juhos et al. (2021) who reported a strong linear correlation between CEC and microbial activity. According to Juhos et al. (2021), linear correlation between CEC and microbial activity is governed by soil organic matter. Soils with high organic matter tend to also have high CEC because organic matter components of soil have negatively charged sites on their surface and these sites attract and exchange positively charged cations (Brady & Weil 2022). G. bicolor does well on rich, shallow sandy soils, occasionally on red clays (Brink 2007), whereas C. apiculatum commonly occurs in sandy soils but disappears completely on clayey soils (Gertenbach 1983). In the opinion of Brady and Weil (2022), increasing the clay content of a sandy soil will help increase its CEC. However, despite C. apiculatum not being able to grow in clayey patches on the crest of the Nkuhlu exclosures in the Kruger National Park, soil samples collected at a radial distance from its canopy had much higher CEC than corresponding G. bicolor.

Our results show that the TN, TC, available P and exchangeable Ca and K of the savanna soils were much higher under trees than at the canopy edge and open grassland zones. However, there were no differences between soil samples collected on the edge of canopy, and these collected outside the tree canopy. Several other researchers have reported similar higher values in soil properties under tree canopies than in open grassland as found in this study (Aweto & Dikinya 2003; Holdo & Mack 2015; Isichei & Moughalu 1992; Malongweni & Van Tol 2022). Higher TN, TC, available P and exchangeable Ca and K under the canopies of the two tree species is presumably because of the effect of litter addition to soil under the canopies, which has resulted in improved fertility in soil under the canopies relative to levels in soil outside the canopies (Aweto & Dikinya 2003; Isichei & Moughalu 1992). It could also be a consequence of microclimatic conditions under tree crowns, such as lower temperatures, reduced light availability and higher soil moisture or combinations of the above. The consequence of temperature reductions under tree canopies is a slower rate of mineralisation (Holdo & Mack 2014; Mordelet et at. 1993). Moreover, reductions in light availability through shade often increase soil nutrient concentrations (Ludwig et al. 2008). Also, as soil samples under the canopy were collected besides the tree trunk, the higher concentration of tree roots near the base of the trees may have had the effect of promoting the build-up of nutrients in that area (Holdo & Mack 2014; Isichei & Moughalu 1992; Malongweni & Van Tol 2022). Tree roots naturally decay on their own thus resulting in improved soil nutrition. Additionally, leachates from tree canopies and nutrient transport by tree roots from the rooting zone to tree canopies may also be a source of higher TN and TC under tree canopies (Malongweni & Van Tol 2022).

This study also revealed that there were no differences in pH between samples collected on the canopy edge and those collected outside canopy; however, these two canopy positions had significantly higher pH than the area next to the tree trunk under the tree crown. It is possible that without crown cover, nutrients on the canopy edge and outside canopy were easily leached from the soil during rainfall. This result supports the assertion of Brady and Weil (2022) that high pH is often caused by excess leaching. Furthermore, since the soil on the edge and outside the tree canopies dries out more because it is exposed to direct solar radiation and this results in the shrinkage of clay colloids, thereby increasing soil pH (Aweto & Dikinya 2003).

The interactive effects of exclosure, woody species and sampling zone had a significant effect on all studied properties except exchangeable Na. Areas under canopy for both G. bicolor and C. apiculatum species in the open access area resulted in significant increases in TC as well as available P. These results reinforce previous findings from other studies as well as our own work on canopy-herbivory interactions in Kruger National Park (Holdo & Mack 2014; Ludwig et al. 2004; Malongweni & Van Tol 2022). Increases in TC and available P caused by the three-way interaction between herbivores, woody species and sampling zone could be primarily caused by herbivore excreta as well as the effect of tree litter on soil organic matter (Holdo & Mack 2014; Malongweni & Van Tol 2022). Organic matter production by litterfall and animal waste and its slow rate of mineralisation under canopies is because of reductions in temperatures there (Isichei & Moughalu 1992). It could also be because of direct biotic effects of trees such as nutrient transport by plant roots from rooting zone to tree canopies (Schroder 2021).

All the trees in the full exclosure had higher pH than those in the open access area, with the canopy edge having the highest pH than other sampling zones. These results contradict the assertion of Ndlela and Schimidt (2016) that mammalian herbivores cause an increase in soil pH of savannas because their excreta, particularly urine, is alkaline. In our study, the reason why the presence of herbivores resulted in a significant decrease in soil pH within the open access area is unclear and we found no evidence to support it.

We also found that G. bicolor in the open access area had significantly higher microbial activity than other treatments, with the canopy edge having the highest activity. However, there were no significant differences in microbial activity between soil samples collected under canopies and those collected on the edge of the tree crown. This could be because G. bicolor regrows quickly after defoliation and/or after being consumed by herbivores (Chira & Kinyamario 2009). This leads to the growth of more leaves, with better quality and greater canopy width. This leads to higher nutrient returns and reduced evaporation losses from the crown-covered surface, which then make the soil prone to microbial metabolism (Chira & Kinyamario 2009; Isichei & Moughalu 1992; Schroder 2021).

In the open access area, the sampling zone besides the tree trunk of both woody species had higher exchangeable K and Ca compared to the full exclosure (other sampling zones). This could be because of urine and dung deposition into the soil from grazing herds of herbivores that congregate under trees. Herbivore excreta improves exchangeable K and Ca by increasing the organic matter content. According to Brady and Weil (2022), organic matter contains several essential nutrients, including K and an increase in soil organic matter content increases the number of colloidal sites for the exchange of cations. This is why from the results of our experiment, it was observed that CEC in the open access area, particularly under tree crown, had much higher CEC than the sampling zones away from canopy and in the full exclosure.