Introduction

Savanna vegetation heterogeneity is shaped by both biotic and abiotic factors which operate at different spatio-temporal scales (Scholes and Archer 1997; Du Toit et al. 2003; Sankaran et al. 2005). At global scale vegetation heterogeneity (composition and structure) is mainly determined by climate (temperature and precipitation) (Sankaran et al. 2005), with fire and herbivory becoming more important at regional to landscape scales (Bond et al. 2005; Asner et al. 2009). Geological variation was observed to be an important source of vegetation heterogeneity within landscapes (Grant and Scholes 2006; Muvengwi et al. 2017) whereas fine spatial scale vegetation heterogeneity has been attributed to dung beetles (Coleoptera) (Andresen 2003), latrines (Pouvelle et al. 2016), beavers (Castor fiber) (Wright et al. 2016), termites (Isoptera) (Holdo and McDowell 2004; Muvengwi et al. 2014) and abandoned kraal sites (van der Waal et al. 2011; Porensky and Veblen 2015).

In African savanna ecosystems, the periodic conversion of cattle (Bos taurus) ranches and livestock pastures into conservation areas has left persistently nutrient-rich patches enriched with dung and urine. Plant productivity and forage quality have been reported to be higher at former kraal sites compared to the expansive landscapes where they occur (Veblen 2012). Hence, kraal sites attract foraging large mammalian herbivores that in turn maintain high nutrient levels through a positive feedback loop when they drop their dung and urine (Augustine et al. 2003; van der Waal et al. 2011). However, some studies in the savannas contended the general observation that nutrient-rich patches are always foraging hotspots for herbivores (Muvengwi et al. 2013; van der Plas et al. 2013), an observation that is attributed to environmental context, where fewer differences in soil nutrients between the nutrient hotspots and the surrounding environment may occur (O’Connor 2013). This observation calls for more empirical studies before generalisations of the effect of nutrient hotspots are made.

Over the last two decades, considerable research effort has been focused on abandoned cattle enclosures, with the aim of understanding how they influence soil, large mammalian herbivory and plant species diversity (Young et al. 1998; Muchiru et al. 2009; van der Waal et al. 2011; Riginos et al. 2012; Veblen 2012; Porensky and Veblen 2015). Soil nutrients have been demonstrated to have a strong positive relationship with forage quality (Holdo and McDowell 2004; Grant and Scholes 2006). For instance, several studies reported twice as high foliar nitrogen and phosphorus on nutrient-rich sites compared with the surrounding savanna landscape (Augustine et al. 2003; Muchiru et al. 2009; van der Waal et al. 2011; Veblen 2012). This has been suggested as a major factor why large herbivores preferentially graze on nutrient-rich sites (Augustine et al. 2003). For example, impala (Aepyceros melampus) and blue wildebeest (Connochaetes taurinus) grazed more on sodic sites (Grant and Scholes 2006), zebra (Equus quagga) and waterbuck (Kobus ellipsiprymnus) on termite mounds (Mobaek et al. 2005) and elephants (Loxodonta africana) on fertilised mopane (Colophospermum mopane) plots (Pretorius et al. 2011).

Different foraging habits and feed preference of large herbivores may influence the dynamics among plant species assemblages (van der Waal et al. 2011; Joseph et al. 2013). During feeding, herbivores tend to select the most conspicuous and palatable plant species thereby causing contrasting effects on growth and dominance of forage species (Augustine and McNaughton 1998; Adler et al. 2001). It has been hypothesised that grasses are more preferred than woody or forbs on nutrient-rich sites despite yielding lesser nutrient content than the later owing to the defensive compounds found in most forb and woody species (Owen-smith and Cooper 1987; Du Toit et al. 2003; van der Waal 2010). To date, several studies have attempted to demonstrate the relationship between forage properties and herbivory on abandoned kraals (reviewed in Riginos et al. 2012; Porensky and Veblen 2015) but with varying findings owing to different environmental contexts and site age.

In order to understand the effect of abandoned kraal sites on herbivory, forage quality and vegetation structure and composition, we sampled vegetation growing on kraal sites and the surrounding savanna matrix (hereafter control plots). The specific objectives of the study were to: (i) test whether soil enriched by cattle dung and urine (during cattle ranching era) and later by soil–plant–herbivore feedback loop (van der Waal et al. 2011) is still detectable in soil after 22 years of abandonment, (ii) assess whether abandoned kraals as sources of habitat scale heterogeneity are influencing vegetation structure and composition, (iii) determine whether abandoned kraals (22 years later) are still areas of high plant productivity and provide nutrient-rich forage, and (iv) assess the relationship between grazing by large mammalian herbivores and forage quality, structure and composition. We hypothesised that abandoned kraals are still nutrient-rich patches affecting vegetation structure and composition and ultimately large mammalian herbivory.

Materials and methods

Study area

The study was conducted in the Save Valley Conservancy, south-eastern lowveld of Zimbabwe (20° 05′S, 32° 00′E) which covers an area of about 3387 km2. Since 1925, the area was used for large scale cattle ranching with an estimated maximum herd of 24,000 cattle in 1975 (Lindsey et al. 2008). The cattle ranching was heavily affected by cattle theft which reduced the herd to 5000 in 1979 and the 1991–1992 drought which forced ranchers to sell all the cattle (Lindsey et al. 2008). By 1992, cattle ranching was abandoned and business shifted to wildlife conservation (Waterhouse 1994). The area lies in the savanna semi-arid zone with annual rainfall ranging from 300 to 500 mm (Lindsey et al. 2008). The geology consists largely of gneisses and paragneisses forming gently, undulating terrain with scattered kopjes. Vegetation is mainly deciduous open woodland savanna with four major woodlands types which are kopje, riverine thicket, mopane and Combretum open and scrub mopane (Waterhouse 1994). Herbaceous community is dominated by a mixture of annual and perennial grasses such as Setaria, Panicum and Cenchrus species. Among its diverse wildlife species, there is: elephant, buffalo (Syncerus caffer), leopard (Panthera pardus), lion (Panthera leo), black rhinocerous (Diceros bicornis), white rhinocerous (Ceratotherium simum) and other ungulate herbivores including Burchell’s zebra, giraffe (Giraffa camelopardalis), impala, sable antelope (Hippotragus niger) and hippopotamus (Hippopotamus amphibius).

Selection of sampling plots

Twelve kraals, abandoned for 22 years (year 1992–2014) were chosen as sampling plots. All the kraals were found with a measurement of 100 × 100 m (a standard kraal size of the then Devuli cattle ranch). Each kraal was used to accommodate about 400 beasts during the night. To correctly identify the locations of the kraals, we engaged a former cattle herder who worked in the conservancy during the cattle ranching era. Remnant features like water troughs, fencing posts and management roads helped us confirm the exact location of the kraals. Similar-sized control plots were marked 200 m from the mid-point of each selected kraal at the same topographical level minimising soil and topography variations following a protocol used by van der Waal et al. (2011). Sampling was done in the wet season (November 2014–April 2015) when savanna vegetation has leaves and easy to identify (Walker 1976).

Soil sampling and analyses

In each sampling plot, five samples were randomly collected using a soil auger to a depth of 15 cm (the herbaceous vegetation primary rooting zone) and thoroughly mixed to obtain a composite sample. Soil samples were chemically analysed for available phosphorus (P), mineral nitrogen (N), extractable potassium (K), exchangeable calcium (Ca), sodium (Na) and magnesium (Mg), organic carbon (C) and pH. Analysis of the soil properties was based on standard laboratory techniques on soil analyses (Anderson and Ingram 1993; Motsara and Roy 2008). Mineral N was determined by micro-Kjeldahl digestion followed by distillation, available P was measured spectrophotometrically following the Olsen’s method and extractable K by a flame photometry (Motsara and Roy 2008). Exchangeable cations (Ca, Mg and Na) were extracted with ammonium acetate and then measured using atomic absorption spectrophotometry (Anderson and Ingram 1993). Soil pH was measured using water and CaCl2 suspension and Organic C was determined volumetrically following procedures described by Motsara and Roy (2008). All laboratory soil analyses were conducted at the Chemistry and Soil Research Institute, Ministry of Agriculture, Mechanisation and Irrigation Development, Harare, Zimbabwe.

Vegetation sampling

Woody vegetation structure and composition was assessed in 100 × 100 m plots following the same procedures described in Gandiwa et al. (2011). For herbaceous vegetation structure and composition, quadrats (1 × 1 m) were placed (10 m apart) along each of the nine selected 100 m transects per plot. All plants encountered were identified to species level and recorded against their relative abundances.

Herbaceous biomass (aboveground) was determined for each sampling plot by measuring the compressed height with a disk pasture meter (calibrated for Lowveld vegetation) dropped twice in every quadrat. Biomass was then calculated using the same equation of Trollope (1990) as follows:

$$Biomass \, \left( {kg \, ha^{ - 1} } \right) \, = \, \left( {\surd X \, \times \, 2260} \right) \, {-} \, 3019,$$

where X is the average disk height for the sampling plot.

Basal and aerial cover estimations followed the 8-point scale by Walker (1976) and every species found within the quadrat was recorded along with the sum cover of that species (Manier and Hobbs 2007).

The level of grazing was assessed using the degree of defoliation of herbaceous tufts (Grant and Scholes 2006). Tuft utilisation assessment was done for every species found in a quadrat and an estimated grazing percentage was recorded. Care was taken not to account for defoliation not related to grazing. Percentage tuft utilisation was calculated from the ratio of utilised tufts over total number of tufts (Grant and Scholes 2006).

Foliar nutrient analyses were done for the commonly found plant species in paired kraal-control plots. Leafy material of commonly found species (trees: mopane, Grewia bicolor and Acacia tortilis; forbs: Sida alba, Solanum incanum and Vernonia myriantha and grasses: Urochloa mosambicensis, Panicum maximum and Chloris virgata) was randomly collected and oven-dried at 70 °C and milled with a 0.5 mm sieve. Chemical analyses were done for nitrogen (N), phosphorus (P), potassium (K), magnesium (Mg), calcium (Ca), iron (Fe), manganese (Mn), zinc (Zn) and boron (B) using standard procedures described in Motsara and Roy (2008).

Data analyses

Data were first tested for normality using Shapiro–Wilk test in the Paleontological Statistics Software package (PAST) (Hammer 2009) and all non-normal, percentage and concentration data were transformed as necessary to improve data distribution. All comparisons between abandoned kraals and controls were done using paired t tests in Statistical Package for Social Scientist (SPSS) version 21 (IBM Corporation 2012). Shannon’s diversity index (H ′), Whittaker’s beta diversity (β w ) and Sørensen’s similarity index were assessed at plot level (Nishizawa et al. 2016). The relationship between levels of grazing and forage quality, as well as the relationship between forage quality and soil conditions was analysed using Pearson’s correlation analysis. Further, we performed simple linear regression analysis to determine the relationship between soil conditions (dependant variables) and plant diversity indices (dependant variables) following Huang et al. (2013). To explore how environmental factors (grazing, foliar and soil nutrient concentration and herbaceous productivity) influence grass and forb species composition, we used multivariate ordination analyses. A constrained ordination Canonical Correspondence Analysis (CCA) with the CANOCO 4.5 package was applied (Lepš and Smilauer 2003).The unimodal CCA was selected because the first axis of the Detrended Correspondence Analysis (DCA) of the data set had a gradient length equal to 5.64 standard deviation units (Lepš and Smilauer 2003). Infrequent species (with abundances < 0.5%) were eliminated and abundances were log10 transformed (Lepš and Šmilauer 1999). Significance of axes and environmental variables were tested by Monte Carlo tests (999 permutations) and ordination diagrams were constructed for the first and second axes.

Results

Influence of abandoned kraals on soil characteristics

Concentrations of soil mineral N, available P, K and extractable Ca were significantly higher on abandoned kraals with factors 1.73, 1.44, 2.59 and 1.32, respectively. Soil pH was more slightly acidic on control plots than abandoned kraals (P = 0.012). Although location showed no significant effect on concentrations of extractable Mg, Na and organic C, concentrations were on average higher on abandoned kraals (Table 1).

Table 1 Comparisons of soil nutrients between abandoned kraals and control plots. (mean ± SE, df = 11)
Full size table

Influence of abandoned kraals on vegetation community structure and composition

A total of 34, 24 and 20 woody, grass and forb species were recorded respectively. Most of the woody species recorded (68%) were found on control plots. Among the woody species, A. tortilis was mostly found (88%) on disturbed kraal sites. Grass species were dominated by U. mosambicensis with 85% found on kraals. Commelina benghalensis was the dominant forb species with 83% found on kraals (Table 2).

Table 2 Number of major plant species observed on abandoned kraals and on control plots (n = 24) in Save Valley Conservancy, Zimbabwe
Full size table

No significant differences in species diversity (Shannon–Wiener index) for woody (P = 0.136), forbs (P = 0.324) and grasses (P = 0.169). However, species richness of woody plants and evenness of forbs differed significantly between abandoned kraals and control plots with the higher values recorded for controls. Vegetation cover (basal and aerial) for both forbs and grasses was significantly higher on abandoned kraals than control plots. Although woody structure was not significantly different between sites, density was significantly higher (P = 0.001) on control plots. In terms of among-site community assemblages, Whittaker’s (β w ) was significantly higher (P = 0.031) for forbs on kraal sites (Table 3).

Table 3 Comparisons of vegetation community composition and structural attributes between abandoned kraals and control plots (mean ± SE, df = 11)
Full size table

Influence of abandoned kraals on herbaceous productivity and herbivory

Abandoned kraals yielded a significantly higher (P = 0.036) herbaceous biomass (aboveground) than the controls (Fig. 1a). Grasses on abandoned kraals were significantly grazed (P = 0.025) than on control plots while grazing percentage for forb species was not significantly different (P = 0.082) between sites (Fig. 1b).

Fig. 1
figure 1

Herbaceous biomass (aboveground) (a) and grazing percentage estimation (mean ± SE) (b) for forb and grass species between abandoned kraals and controls in Save Valley Conservancy, Zimbabwe

Full size image

Influence of abandoned kraals on foliar nutrient concentration

Abandoned kraals supported grasses with significantly higher concentrations of foliar N, P, K, Ca, Mg, Fe, Mn, Zn and B with factors 1.21, 1.67, 1.28, 1.29, 1.43, 1.23, 1.42, 1.70 and 1.49, respectively (Fig. 2a–i).

Fig. 2
figure 2

A kraal-control comparison of foliar nitrogen (a), phosphorus (b), potassium (c), calcium (d), magnesium (e), iron (f), manganese (g), zinc (h) and boron (i) (mean ± SE) of commonly found woody species (Colophospermum mopane, Grewia bicolor and Acacia tortilis), forbs (Sida alba, Solanum incanum and Vernonia myriantha) and grasses (Urochloa mosambicensis, Panicum maximum and Chloris virgata) in Save Valley Conservancy, Zimbabwe

Full size image

Unlike grasses, forbs showed a similar concentrations of foliar nutrients with the exception of K which was 1.33 times higher (P = 0.023) on abandoned kraals (Fig. 2c). Woody species found on abandoned kraals had higher (P < 0.05) concentrations of foliar K, Mg, Mn and B with factors 1.33, 1.53, 1.36 and 1.31, respectively (Fig. 2c, e, g and i).

Relationship between forage species community to environmental factors

Grasses

The results of the CCA showed that the first axis accounted for 47% of the variance. The species-environment correlations for the two axes were high (Table 4). The first axis explained 32.9% of the variation (Monte Carlo permutation test, F-ratio = 2.17, P = 0.038). The permutation test showed that biomass explains a significant proportion of the species assemblages. Also, the small angles between the arrows for N, P, K, Gr % (grazing percentage) and Bm (biomass) indicate a positive correlation between those variables and forage species. Explanatory variables; BC, SW, Gr % and P showed no significant influence on the structure and composition of forage species (Fig. 3; Table 4).

Table 4 Ordination CCA results for the first two axes for grasses (Fig. 3) and forbs (Fig. S1) and environmental variables on abandoned kraals and control plots in Save Valley Conservancy, Zimbabwe
Full size table
Fig. 3
figure 3

Canonical Correspondence Analysis triplot for grass species and environmental variables from abandoned kraals and control plots in Save Valley Conservancy, Zimbabwe. Environmental variables: N Nitrogen, P Phosphorus, K: Potassium, BC Basal cover, Bm Biomass, Gr % grazing percentage and SW: Shannon–Wiener diversity. Grass species: E. colon Eragrostis colona, D. velut Digitaria velutina, E. visco Eragrostis viscosa, E. barre Eragrostis barrelieri, S. sagit Setaria sagittifolia, U. brach Urochloa brachyuran, E. cylin Eragrostis cylindrifolia, E. trico Eragrostis tricophora, D. gigan Dactyloctenium gigangteum, C. virga Chloris virgata, T. racem Tragus racemosa, U.mos Urochloa mosambicensis, S. stapf Sporobolus stapfianus, P. max Panicum maximum, D. erianth Digitaria eriantha, P. dist Panicum distichum, T. barte Tragus barteronianus, A.conge Aristida congesta, C. roxb Chloris roxburghiana, A. sciur Aristida sciurus, P. natal Panicum natalense, E. rigidior Eragrostis rigidior, A. adsce: Aristida adscensionis and S. panic Sporobolus panaicoides

Full size image

Forbs

The first two axes for the CCA for forbs (F-ratio = 1.79, P = 0.276) and all other axes combined (F-ratio = 1.03, P = 0.368) explained the variation of the data although they were not statistically different (Table 4). Although no clear separation of samples (kraals and controls) was observed, all the axes had species-environment correlations greater than 0.6 (Fig. S1). The first axis had a moderately strong negative species-environment correlation (r = − 0.686) and represents a gradient of decreasing diversity (SW) associated with B. pilosa (highest score). The second axis represent a gradient of increasing basal cover (r = 0.682) with an association with A. hispidium. All explanatory variables had a non-significant relationship with forb species and biomass while soil N, P and K had very weak correlations with both axes (Table 4).

Soil nutrient concentrations showed weak correlations with diversity and richness of woody, grass and forb species (Table 5). Regression analyses between major soil nutrients (N, P and K) and diversity indices showed no significant relationships for both kraals and control plots (Fig. S2). Level of grazing also had a weak relationship with richness and diversity of grass and forb species (Table 5). Foliar P showed significant positive correlations with soil Mg (r = 0.864) and organic C (r = 0.703). Foliar Mg (r = 0.774) and Mn (r = 0.753) showed significantly positive correlations with level of grazing for forb species (Table 6).

Table 5 Pearson’s correlation coefficients (r) between environmental variables (soil quality and level of grazing) and diversity measures (Shannon diversity H′ and species richness, S) of woody, grass and forb species in Save Valley Conservancy, Zimbabwe (n = 24)
Full size table
Table 6 Pearson’s correlation coefficients (r) between forage and soil quality factors and level of grazing for forb and grass species in Save Valley Conservancy (n = 24)
Full size table

Discussion

Our results showed that abandoned kraals were still nutrient-rich after 22 years of abandonment with a high production of nutritious grass plants preferred by herbivores. Surprisingly, among the forage species, the dominant forb species on abandoned kraals were least consumed by herbivores and showed higher foliar nutrient concentrations between kraals and controls. Contrary to the present study, plant diversity did not respond to abandoned kraals and plant cover was less important in explaining the plant community composition. Plant diversity had a negative relationship with grazing and forage quality.

Soil nutrient levels were higher on abandoned kraals than control plots suggesting the persistence of nutrients a long time after abandonment. These findings corroborate other studies done on abandoned kraals in the savannas (Kizza et al. 2010; van der Waal et al. 2011; Porensky et al. 2013). Soil nutrients on nutrient hotspots has been reported to be largely maintained by a positive feedback loop whereby herbivores feeding on them drop dung and urine in the process further enriching the sites (Augustine and McNaughton 2006; van der Waal et al. 2011). For instance, in this study the limiting soil nutrients N, P, K, and Ca were much higher on abandoned kraals than control plots.

Previous studies have demonstrated that grazing on nutrient-rich patches may alter the trends in plant community structure and composition (Manier and Hobbs 2007; Riginos et al. 2012; Young et al. 2013). We recorded that nutrient-rich forage found on abandoned kraals and the subsequent increased grazing were not associated with changes in alpha and beta diversity of grass, forb and woody species. However, our results must be treated with caution since we did not use exclosure plots on kraals. These findings are contrary to the productivity-diversity hypothesis whereby plant diversity was supposed to be lower on nutrient-rich abandoned kraals likely due to the competitive exclusion by fewer dominant species (Grime 1973). Other studies have concluded that plant diversity increases with high grazing on nutrient-rich habitats and decreases in nutrient-poor habitats (see Okullo and Moe 2012). According to the grazer-reversal hypothesis (Proulx and Mazumder 1998), diversity increases with high grazing on nutrient-rich habitats and decreases under nutrient-poor conditions. A lower plant diversity on nutrient-rich habitats compared to the surrounding matrix has been recorded by others in the savannas (Veblen and Young 2010; Okullo and Moe 2012; Porensky et al. 2013; Porensky and Veblen 2015) suggesting the dominance of a fewer competitive forb species over grasses on nutrient-rich areas. We also suggest that after two decades plant species mix could have been achieved through seed dispersal via ungulates and birds.

A lower woody species density on abandoned kraals points to the constrained establishment of woody species in a grass dominated habitat (Scholes and Archer 1997; van der Waal 2010) and probably the destructive mechanical clearance of vegetation done during setting up of kraals. However, woody height, basal area and canopy volume were similar between sites mainly because their means were skewed by extreme values of remnant large trees left inside kraals. An increase in productivity of herbaceous plants (grasses and forbs) on abandoned kraals can be attributed to their competitive growth advantage over woody species on areas of high soil fertility and increased sprouting of annuals during the wet-season (van der Waal 2010). Elsewhere, forbs and graminoids have been reported to dominate nutrient-rich habitats such as termite mounds (Okullo et al. 2013) and abandoned pastoral settlements (Muchiru et al. 2009).

Herbaceous biomass and plant cover were twice as high on abandoned kraals perhaps showing the positive response of herbaceous plants to limiting soil nutrients in semi-arid savannas (Veblen and Young 2010). A higher concentration of foliar nutrients for grasses on abandoned kraals compared to the same species in the matrix, suggest why herbivory was high on abandoned kraals. According to optimal foraging strategies, herbivory is much influenced by the quality and quantity of forage (Owen-smith and Cooper 1987; Grant and Scholes 2006). Regardless the smaller size of kraals (100 × 100 m) compared to the matrix, herbivores preferred feeding on these sites perhaps because they supported high quantity and nutritious forage. Perhaps, forb species were not as much preferred as grass species on abandoned kraals probably because of their greater chemical defence or lower nutritional returns to animals (Ball et al. 2001; Du Toit et al. 2003). This result is in line with the findings of Riginos and Grace, (2008) who recorded a decreased consumption of forbs despite higher protein levels than grasses.

Results from the CCA for grasses indicated that herbaceous biomass is an important factor explaining the differences in plant structure and composition between abandoned kraals and control plots. However, others have demonstrated that nutrient-rich sites are usually dominated by plant monocultures which yield a high biomass per unit area (Laossi et al. 2008; Veblen 2012). Considering CCA results for forbs, no obvious pattern was evident between measured explanatory variables and forb species at both sites pointing to the unresponsive nature of forbs to soil fertility levels and poor grazing preference by herbivores (Ball et al. 2001; Du Toit et al. 2003). In contrast to our study hypothesis, results showed that grazing had a non-significant relationship with grass communities whereas species diversity had a negative relationship with both grass and forb communities. This result might be explained by the existence of most palatable and acceptable grasses recorded in this study which could perhaps reduce feed selectivity by herbivores (Augustine and McNaughton 1998). It is important to note that unlike the highly preferred grass species, forb species respond differently to herbivory on abandoned kraals.

Conclusion

This study adds to the current available knowledge of an ecosystem in a previously disturbed conservation area which has been left with remnant nutrient-rich patches. Residual soil nutrients were still higher on kraals, perhaps being maintained by plant-herbivore-soil nutrient feedbacks. Abandoned kraals harbour grass forage of high quality which is more utilised by herbivores. Utilisation of forage was more explained by foliar nutrient concentrations and herbaceous productivity than plant diversity. Among the forage species, forbs had low foliar nutrient concentration and were less preferred by herbivores at least in the rainy season. Future studies should therefore explore the spatio-temporal links between herbivory of forbs and graminoids on nutrient-hotspots in semi-arid savannas and elsewhere in an exclusion experiment. We recommend a conservation monitoring programme on abandoned kraals as they are serving as nutrient-hotspots for herbivores in an otherwise dystrophic semi-arid savanna.