Savanna ecosystems are dominated by two distinct plant life forms, grasses and trees, but the interactions between them are poorly understood. Here, we quantified the effects of isolated savanna trees on grass biomass as a function of distance from the base of the tree and tree height, across a precipitation gradient in the Kruger National Park, South Africa. Our results suggest that mean annual precipitation (MAP) mediates the nature of tree-grass interactions in these ecosystems, with the impact of trees on grass biomass shifting qualitatively between 550 and 737 mm MAP. Tree effects on grass biomass were facilitative in drier sites (MAP≤550 mm), with higher grass biomass observed beneath tree canopies than outside. In contrast, at the wettest site (MAP = 737 mm), grass biomass did not differ significantly beneath and outside tree canopies. Within this overall precipitation-driven pattern, tree height had positive effect on sub-canopy grass biomass at some sites, but these effects were weak and not consistent across the rainfall gradient. For a more synthetic understanding of tree-grass interactions in savannas, future studies should focus on isolating the different mechanisms by which trees influence grass biomass, both positively and negatively, and elucidate how their relative strengths change over broad environmental gradients.
Citation: Moustakas A, Kunin WE, Cameron TC, Sankaran M (2013) Facilitation or Competition? Tree Effects on Grass Biomass across a Precipitation Gradient. PLoS ONE 8(2): e57025. https://doi.org/10.1371/journal.pone.0057025
Editor: Harald Auge, Helmholtz Centre for Environmental Research – UFZ, Germany
Received: June 6, 2012; Accepted: January 21, 2013; Published: February 22, 2013
Copyright: © 2013 Moustakas et al. This is an open-access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.
Funding: This research was funded by a NERC Research Grant to MS, WEK, and AM (NE-E017436-1). TCC was funded by the University of Leeds. MS would also like to acknowledge support provided by the Ramalingaswami Re-entry Fellowship, Department of Biotechnology, Government of India. AM thanks the School of Biological and Chemical Sciences, Queen Mary University of London and the Head of School, Matthew Evans for covering the publications fees. The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript.
Competing interests: The authors have declared that no competing interests exist.
Savannas are ecosystems characterised by a continuous grass layer and a discontinuous tree layer. They cover over 20% of the Earth's total land surface and support a significant proportion of the planet's livestock and wild herbivores , . The ratio of trees to grasses in savannas can vary depending on several environmental factors with precipitation and soil properties generally considered the predominant drivers at large scales , , , ,  which in turn modulate plant-plant interactions at local scales , , .
Traditionally, ecologists have emphasized the role of competition between trees and grasses as being a key determinant of savanna structure , , . While the importance of competition in structuring ecological communities is widely recognized, there is also a growing appreciation of the role of facilitation amongst plants in structuring ecological communities , , , , especially in stressful environments , ; indeed facilitation is a process that needs to be more integrated into ecological theory . Facilitation can occur through various mechanisms including refuge from physical stress , refuge from predation , refuge from competition , and improved resource availability .
At present, the importance of tree-grass facilitation relative to competition, and the role of microhabitats and microclimatic factors created by the presence or absence of trees for grasses in savannas has not yet been fully explored (but see , , , , , ). While it is well established that trees in savannas compete with grasses for light, nutrients, and water , , there are also several cases where trees have been reported to have facilitative effects on grasses , , , . For example, grass biomass has been found to be higher under tree canopies when compared to the interspaces between trees in several systems , , , , , . However, grass biomass has also been reported to be lower beneath tree canopies in some savannas , , , while other studies have found no differences in grass biomass beneath tree canopies and in tree interspaces . At present, the reasons underlying these divergent responses are unclear.
Savanna trees can facilitate grasses by altering resource availability and microclimatic conditions, and providing grasses refuges from grazing. Trees in particular affect water redistribution in the landscape , and play important roles by creating shade , ,  and by drawing water from deep sources inaccessible to grasses, i.e. hydraulic lift , , . However, the extent to which such positive effects offset the negative effects of competition is unclear. For example, in a study of hydraulic lift (the process of water movement from relatively wet to dry soil layers through plant roots) of Acacia tortilis it was found that the δ18O of water extracted from the xylem water of grasses indicated that when they grew near trees they had values similar to those of groundwater either because grasses could use water from deeper soils or because they used water hydraulically lifted by trees . However, at the same site  found lower soil moisture content under trees than in the open, both during dry and wet seasons, and marginally higher grass biomass in open areas. Thus, while hydraulic lift did facilitate water uptake by grasses, the effects of competition with tree roots cancelled the beneficial influence of tree roots on grass biomass at that site . Overall, findings in savannas have shown that both facilitation and competition can occur in the same ecosystem, and that competitive and facilitative effects do not always balance .
In this study, we examined the effects of isolated trees on grass biomass across a precipitation gradient in an African savanna. Specifically, we looked at how grass biomass changed as a function of distance from the base of the tree and with tree size, and how this relationship was influenced by precipitation. In arid and semi-arid savannas where water is the main limiting resource , , we expected trees to facilitate grasses by enhancing water availability, and predicted that grass biomass values would be higher in the sub-canopy areas than in tree interspaces. In contrast, in more mesic savannas, where water is typically less limiting and factors such as shading by trees becomes increasingly important, we expected grass biomass to be lower in sub-canopy areas than in the interspaces between trees. Further, since the microclimate experienced by grasses is also likely to be affected by individual tree characteristics such as size, we expected that the extent to which trees facilitate or compete with grasses would change with tree size. Increases in tree size can lead to increased soil resource availability and hence increased sub-canopy grass biomass as a result of hydraulic lift or increased nutrient contents below canopies . Alternately, increases in tree size can also result in increased solar radiation and evapotranspiration in sub-canopy areas leading to lowered soil moisture and sub-canopy grass biomass , . We expected sub-canopy grass biomass to increase with tree size with such effects particularly pronounced in mesic areas where sub-canopy grasses tend to be light-limited, and where increased solar irradiation beneath larger trees can in fact have a positive effect on sub-canopy grass biomass.
The study was conducted in the Kruger National Park (KNP), South Africa between January and February 2008. The park is situated in the savannas of north-eastern South Africa, and covers an area of ∼19,633 km2. Altitude ranges from 260 to 839 m above sea level within the park. Mean Annual Precipitation (MAP) varies from around 750 mm in the south to approximately 350 mm in the north, with marked annual variations . The vegetation in the park is characterized by dense savanna with species such as Acacia nigrescens, Sclerocarya birrea, Combretum imberbe, Colophospermum mopane, Terminalia sericea and Burkea africana dominating the canopy depending on the location within the park .
Our study was conducted in the long-term Experimental Burn Plots (EBPs) of the Kruger National Park . Only ‘unburnt’ treatment plots were sampled (i.e. fire exclusion plots) for this study. Plots were located in four major landscapes of the park underlain by both granites: Pretoriuskop (MAP = 737 mm) and Skukuza (MAP = 550 mm), and basalts: Satara (MAP = 544 mm) and Mopane (MAP = 496 mm). In each of the four landscapes, there were four replicate plots, each covering an area of ∼7ha. Fire had been excluded from our study plots for more than 50 years , . All necessary permits for field work were obtained from the Administration, Scientific Services and the local Rangers of the Kruger National Park.
We identified isolated trees in each of the replicate plots (N = 16; four in each landscape). Isolated trees were defined as those for which distance to the nearest woody plant (tree or shrub) neighbour was as least three times the canopy radius of the focal tree. For each tree, we recorded height, girth at breast height, and canopy diameter along two perpendicular axes. In cases where measuring girth at breast height was not feasible we recorded their girth at the closest available point. In addition, we also measured grass biomass at different distances from the base of the tree, corresponding to 50%, 100% (i.e. canopy edge or drip line), 150% and 200% of the tree canopy radius . Distances were measured as a proportion of canopy size rather than as absolute values in order to facilitate comparison of canopy effects on grass biomass of uneven-sized trees . For each tree, grass biomass was measured along 3 transects radiating away from the base of tree, each 1200 apart from the other. Grass biomass measurements were taken at peak herbaceous standing crop in January and February. Biomass values from the three transects were averaged to get a mean value for grass biomass for each distance category for each tree. Grass biomass was estimated using a disc pasture meter  specifically calibrated for KNP , . Disc pasture meter calibrations, conducted using samples from across the full extent of KNP, indicate a high degree of concordance between measured and estimated values of grass biomass in KNP (R2 = 0.972, 45). Grass biomass was estimated from disk pasture meter readings (in cm) using the formula y = −301.9+226, where x is the disc reading in cm and y the grass biomass in g/m2 . For a more detailed description of the device and its calibrations for KNP see , .
In all, a total of 93 trees were sampled across all plots (Table 1). One plot (Numbi block in Pretoriuskop) had no isolated trees as per our criterion as a result of the dense nature of the vegetation in the plot.
Linear mixed effects models were used to quantify the effects of tree characteristics on grass biomass across sites. Rainfall, distance from the base of the tree, and tree size were the factors included in the analysis (fixed effects), with geology (granite or basalt) included as a random effect in the model. We chose to include geology as a random rather than fixed effect because the design of the Experimental Burn Plots in KNP does not allow separating out the individual and interactive effects of both rainfall and geology simultaneously; the two lower rainfall sites (Mopane and Satara) are on basalts and the two higher rainfall sites (Skukuza and Pretoriuskop) are on granites. Our dataset also contained a different number of species at each site. Because sample sizes were limited, we were not able to explicitly test to see if tree effects on grass biomass varied depending on tree species identity. However, since tree species identity can potentially play a role in regulating tree-grass interactions in savannas , ,  we additionally included tree species identity as a random effect in our analysis. An initial calculation of the contribution of the random structure (Standard Deviations of the random effects from our model) showed that the factor that contributes least to variance in the estimates of the mean grass biomass is geology (Random effects: ∼1|geology, Intercept StdDev = 17.73082; ∼1 | geology/plot, Intercept StdDev = 114.0258; ∼1 | geology/plot/species, Intercept StdDev = 106.3738; ∼1 | geology/plot/species/treeID, Intercept StdDev = 54.58133). A separate evaluation of the contribution of species identity using a linear NULL model, (ANOVA(biomass∼plot/species/treeID) where the nested structure is a fixed effect, showed that there is a greater contribution to the variance in grass biomass from plots within sites (σ2 = 14646) or between individual trees within species (σ2 = 24232) than between tree species (σ2 = 3023).
In the mixed effects model used, grass biomass data were grouped by distance increments within individual trees nested by species within plots, within geology to account for non-independence of data from trees on the same site . Data grouping accounts for autocorrelation between samples in all its forms . Thus, although we only report on the effects of trees on grass biomass independent of tree species identity, our analysis nevertheless accounted for the fact that there are differences in tree species composition across our study sites.
We evaluated the effects of three correlated measures of tree size – height, basal area and canopy area – on grass biomass across sites. We created three different maximal models on non-transformed grass biomass with different combinations of the correlated fixed effects: tree height, canopy area and basal area (as indices of tree size) as well as distance from the base of the tree and site. The maximal models included all possible interaction terms. We used the Akaike Information Criterion (AIC) to assess the best maximal model. Of the three indices of tree size evaluated, the maximal model which included tree height as the index of tree size was the best, and we only report results from this model here. The maximal model which included tree height as the index of tree size was subsequently simplified via stepwise deletion wherein non-significant factors and their interactions were sequentially removed, until further simplification was not justified. Model selection was conducted using the AIC with maximum likelihood estimation, but the presented fit of the minimal model used restricted maximum likelihood (REML); . Any deletion that did not increase AIC scores by more than 2 was deemed to be justified . The minimum adequate model selected by AIC or comparative F-tests were identical. Inspection of residual plots for constancy of variance and heteroscedasticity indicated that the model was well behaved in all cases. All analyses were conducted in R 2.14.1 using the nlme package .
The minimum adequate model explaining grass biomass included the main effects of rainfall, distance and tree height and the two-way interactions between rainfall and distance, and rainfall and height (Table 2).
Tree effects on grass biomass beneath canopies changed across the rainfall gradient. In the three drier sites (Mopane MAP = 494 mm, Satara MAP = 544 mm, & Skukuza MAP = 550 mm), grass biomass was significantly higher beneath tree canopies than outside canopies (Table 2 and Figure 1). In contrast, there were no significant differences in grass biomass beneath and outside canopies at the wettest site (Pretoriuskop, MAP = 737, Fig. 1).
Biomass values at 100% correspond to the canopy edge or drip line. Bars represent ±2 standard errors. Grass biomass is greater beneath the canopy (50% distance) compared to outside the canopy at the three drier sites, but not at the wettest site.
The interaction between rainfall and tree height was marginally significant (P = 0.053) suggesting that effects of tree height on grass biomass differed between sites (rainfall x height interaction, Table 2). At the driest site (Mopane), sub-canopy grass biomass increased as tree height increased (adjusted R2 = 21.1%, p = 0.039, Figure 2a). A similar pattern was observed at Skukuza (adjusted R2 = 9.9%, p = 0.048; Figure 2c). At the wettest site (Pretoriuskop) grass biomass below tree canopies increased with tree height but that relationship was only marginally significant (adjusted R2 = 10.2%, p = 0.055; Figure 2d). At Satara there was no effect of tree height on grass biomass (adjusted R2 = 0%, p = 0.975; Figure 2b). These observed patterns were not a consequence of consistent differences in tree architecture or height resulting from species turnover across sites; tree height distributions were similar across our four study sites (Figure 3a), with taller trees having proportionally larger canopies regardless of species identity (Figure 3b).
Sites are ordered in terms of increasing rainfall from Mopane (a) to Pretoriuskop (d). Sub-canopy grass biomass refers to grass biomass measured at 50% of the canopy radius, i.e. half way between the base of the tree and canopy edge. Solid lines are the best fit regression, dashed lines the 95% confidence interval, and the dotted lines the 95% predicted interval. Removal of the tallest tree in Satara as an outlier does not qualitatively change the outcome of the regression.
The solid line is the median, and the boxes are defined by the upper and lower quartile (25th and 75th percentiles). The whiskers extend up to 1.5 times the inter-quartile range of the data. The figure indicates that distribution of tree heights was not uneven between study sites. Relationship between average canopy diameter (m) and tree height (m) across all sampled trees (b). Average canopy diameter is the mean of canopy diameters measured along two perpendicular axes. Regression results indicate a tight relationship between canopy diameter and tree height, with taller trees having proportionally larger canopies regardless of species identities (adjusted R2 = 53.8%, p<0.0001. CI: 95% confidence interval, PI: 95% predicted interval).
Our results indicate that the nature of tree-grass interactions changes from positive to negative across a gradient of increasing precipitation. We suggest that this change occurs due to a decline in the relative importance of facilitation of grasses by trees, relative to competition between them, with increasing rainfall.
According to our results, the net impact of trees on grass biomass appears to shift qualitatively between 550 (Skukuza) and 737 (Pretoriuskop) mm MAP (Table 1 and Figure 1) in our study site. This is in accordance with the results of previous studies showing that tree effects on grass biomass are more positive on arid sites than in mesic ones , . Grass biomass has been reported to be higher below tree canopies in more arid savannas (MAP<∼650 mm; e.g. , , ), and lower below tree canopies in more mesic sites (MAP>∼650 mm; e.g. , ). In sites with intermediate rainfall (MAP≈650 mm), grass biomass did not appear to be significantly different beneath and away from tree canopies e.g. , .
Our results are also in accordance with previous empirical and theoretical studies of facilitative-competitive interactions from other systems which suggest that the relative importance of facilitation versus competition should vary across gradients of abiotic stress, with facilitation encountered in more stressful environments. , . Facilitation, or positive interactions across plant communities, has been reported across precipitation ,  altitudinal , , temperature , and slope  gradients. Although we have not specifically quantified the mechanisms underlying the observed patterns in our study, we suspect that it is predominantly a result of improved water conditions beneath trees in arid sites. Other studies have argued that facilitative and competitive effects are density-dependent rather than driven by only abiotic factors; low plant densities are favourable for facilitative effects, while increased plant densities favour competition . Of course, the latter is often correlated with abiotic factors, and higher plant densities are usually encountered in areas of lower abiotic stress. In a meta-analysis of plant interactions in arid environments it was found that the effects of neighbours was density-driven with positive net effects of neighbours occurring at low abiotic stress and negative effects at high stress . Likewise, a study conducted at an area of MAP = 500 mm reported that few, isolated trees had positive local effects on savanna grasses, but in areas of high tree density the negative landscape-scale effects of trees outweighed these positive effects . Our results show that switches from facilitation to neutrality or competition can also occur independent of density dependence since our results are based exclusively on isolated trees. A potential explanation for the absence of facilitation in the more mesic sites is that tree Leaf Area Index (LAI) tends to increase with increasing rainfall , , and thus mesic sites are likely to be associated with lower light penetration through tree canopies, with potential negative impacts for shade intolerant C4 grasses.
Within the overall precipitation-driven pattern, our results indicate that understorey grass biomass can be additionally influenced by tree characteristics such as height. For example, tree height was a significant factor influencing grass biomass in our most arid study site, explaining up to 21% of the variance in sub-canopy grass biomass. To the best of our knowledge, very few studies have examined the effect of tree height on tree grass interactions, and those that have report contrasting patterns. In a study examining the relationship between tree height and grass biomass in savannas, no relationship was found between grass biomass, tree height, and distance from the canopy, despite soil nutrient concentrations being much higher under larger trees , , on the other hand, examined the effects of Colophospermum mopane trees on understorey vegetation, and found grass biomass to be higher below tree-canopies, with effects more pronounced under large canopies. Similarly, in a study where tree age was included as a factor potentially mediating tree effects on grass biomass, older trees were found to facilitate grasses more than younger trees, and this was attributed to the fact that older trees had a higher fraction of deep rather than lateral roots . In contrast,  found that taller Acacia karroo trees suppressed grasses more than shorter ones in a semi-arid savanna in South Africa.
Tree size can influence sub-canopy grass biomass by i) altering soil resource availability in sub-canopy areas, ii) modulating solar irradiation and hence evapotranspiration and soil temperatures in sub-canopy areas, and/or iii) regulating access to grazers and thus influencing grass offtake from sub-canopy areas , , , . Increases in tree size can lead to increased soil resource availability and hence increased sub-canopy grass biomass as a result of hydraulic lift or increased nutrient contents below canopies , and such effects may be manifest in both arid and mesic sites. However, increases in tree height can also result in a higher canopy and thus increased solar radiation and evapotranspiration in sub-canopy areas leading to lowered soil moisture and sub-canopy grass biomass, particularly in arid areas where water is the limiting resource. Similarly, in mesic areas where sub-canopy grasses are predominantly light-limited, increased solar irradiation beneath larger trees can in fact have a positive effect on sub-canopy grass biomass. Finally, increased grazing pressure as a result of greater access to sub-canopy grasses beneath taller trees can result in sub-canopy grass biomass decreasing with tree size , , , with such effects more pronounced in arid and semi-arid areas which typically support higher grazer biomass . Ultimately, the net effect of increasing tree size on sub-canopy grass biomass is contingent on the relative strengths of these different processes. The lack of a consistent relationship between tree size and sub-canopy grass biomass in our study suggests that the relative strengths of these different processes varied differentially across the rainfall gradient in our study area.
The design of the long-term experimental burn treatments at KNP, where the drier sites occur on basalts and the wetter sites on granites, did not allow us to explicitly isolate the effects of geology from the larger scale rainfall driven patterns. However, our results indicate that the contribution to the overall variance accounted for by including geology as a random effect was relatively small when compared to that resulting from differences between individual trees within a species, differences between tree species and between replicate plots within a site. Furthermore, the overall pattern of greater grass biomass beneath tree canopies compared to tree interspaces was evident at both Satara and Skukuza, sites which are characterized by similar rainfall but differ in their underlying geologies. Future studies that explicitly sample across comparable, broad rainfall gradients on both granites and basalts, will help determine the extent to which the rainfall driven switch from facilitation to competition is influenced by underlying geology.
Our results indicate that the net effect of savanna trees on grasses is contingent on environmental context, with facilitation outweighing competition in arid sites and competition predominating in more mesic sites. Although our analysis did not examine how the nature of tree-grass interactions changes over time, interactions in plant communities can also switch between positive and negative contingent on temporally-generated gradients at the same location too; for example as a result of seasonal changes in climate, grazing pressure, etc. , . Future studies should focus not only on isolating the different mechanisms by which increases in tree size influence grass biomass and production and how this changes across broad environmental gradients, but also on the extent to which such effects change over time and are dependent on tree species identity. This will provide for a more integrated understanding of tree-grass interactions in savannas.
We thank the Scientific Services, Kruger National Park for providing us all necessary permits to carry out this research. We also thank the Scientific Services and the local rangers of the Kruger National Park for logistical help, and for facilitating this work. We thank Johan Baloyi and Elliot Thekiso Shilote for assistance during fieldwork. We would also like to thank Navashni Govender for all her help, and for providing us the disc pasture meter and information on the Experimental Burn Plots. We also thank Rupert Quinnell and Alexandros Galanidis for their help with statistical analysis. Comments from 2 anonymous reviewers considerably improved an earlier manuscript draft.
Conceived and designed the experiments: AM MS. Performed the experiments: AM. Analyzed the data: AM MS TCC. Wrote the paper: AM MS WEK TCC.
- 1. Scholes RJ, Archer SR (1997) Tree-grass interactions in savannas. Annual Review of Ecology and Systematics 28: 517–544.
- 2. Sankaran M, Hanan NP, Scholes RJ, Ratnam J, Augustine DJ, et al. (2005) Determinants of woody cover in African savannas. Nature 438: 846–849.
- 3. Scholes RJ, Walker BH (1993) An African savanna - synthesis of the Nylsvley study. Cambridge University Press.
- 4. Sankaran M, Ratnam J, Hanan NP (2008) Woody cover in African savannas: the role of resources, fire and herbivory. Global Ecology & Biogeography 17: 236–245.
- 5. Moustakas A, Wiegand K, Ward D, Meyer KM, Sankaran M (2010) Learning new tricks from old trees: revisiting the savanna question. Frontiers of Biogeography 2: 47–53.
- 6. Knoop WT, Walker BH (1985) Interactions of woody and herbaceous vegetation in a southern African savanna. Journal of Ecology 73: 235–253.
- 7. Callaway RM, Nadkarni NM, Mahall BE (1991) Facilitation and interference of Quercus douglasii on understory productivity in Central California. Ecology 72: 1484–1499.
- 8. Callaway RM, Davis FW (1993) Vegetation dynamics, fire, and the physical environment in coastal Central California. Ecology 74: 1567–1578.
- 9. Walter H. (1971) Ecology of tropical and subtropical vegetation. Oliver & Boyd, Edinburgh, UK.
- 10. Sankaran M, Ratnam J, Hanan NP (2004) Tree–grass coexistence in savannas revisited –insights from an examination of assumptions and mechanisms invoked in existing models. Ecology Letters 7: 480–490.
- 11. Brooker RW, Maestre FT, Callaway RM, Lortie CL, Cavieres LA, et al. (2008) Facilitation in plant communities: the past, the present, and the future. Journal of Ecology 96: 18–34.
- 12. Barbier N, Couteron P, Lefevre R, Deblauwe V, Lejeune O (2008) Spatial decoupling of facilitation and competition at the origin of gapped vegetation patterns. Ecology 89: 1521–1531.
- 13. Freckleton RP, Watkinson AR, Rees M (2009) Measuring the importance of competition in plant communities. Journal of Ecology 97: 379–384.
- 14. Bullock JM (2009) A long-term study of the roles facilitation in the establishment following heathland fires. Journal of Ecology 97: 646–656.
- 15. Maestre FT, Valladares F, Reynolds JF (2005) Is the change of plant-plant interactions with abiotic stress predictable? A meta-analysis of field results in arid environments. Journal of Ecology 93: 748–757.
- 16. Maestre FT, Callaway RM, Valladares F, Lortie CJ (2009) Refining the stress-gradient hypothesis for competition and facilitation in plant communities. Journal of Ecology 97: 199–205.
- 17. Bruno JF, Stachowicz JJ, Bertness MD (2003) Inclusion of facilitation into ecological theory. Trends in Ecology & Evolution 18: 119–125.
- 18. Tirado R, Pugnaire FI (2005) Community structure and positive interactions in constraining environments. Oikos 111: 437–444.
- 19. Harmon ME, Franklin JF (1989) Tree seedlings on logs in Picea-Tsuga forests of Oregon and Washington. Ecology 70: 48–59.
- 20. Callaway RM (1995) Positive interactions among plants (Interpreting botanical progress). The Botanical Review 61: 306–349.
- 21. Belsky AJ, Amundson RG, Duxbury JM, Riha SJ, Ali AR, et al. (1989) The effects of trees on their physical, chemical, and biological environment in a semi-arid savanna in Kenya. Journal of Applied Ecology 26: 1005–1024.
- 22. Weltzin JF, McPherson GR (2000) Implications of precipitation redistribution for shifts in temperate savanna ecotones. Ecology 81: 1902–1913.
- 23. Belsky AJ (1994) Influences of trees on savanna productivity: tests of shade, nutrients, and tree-grass competition. Ecology 75: 922–932.
- 24. Weltzin JF, McPherson GR (1997) Spatial and temporal soil moisture partitioning by trees and grasses in a temperate savanna, Arizona, USA. Oecologia 112: 156–164.
- 25. Xu C, Liu M, Zhang M, Chen B, Huang Z, et al. (2011) The spatial pattern of grasses in relation to tree effects in an arid savanna community: Inferring the relative importance of canopy and root effect. Journal of Arid Environments 75: 953–959.
- 26. Riginos C (2009) Grass competition suppresses savanna tree growth across multiple demographic stages. Ecology 90: 335–40.
- 27. Belsky AJ, Mwonga SM, Amundson RG, Duxbury JM, Ali AR (1993) Comparative effects of isolated trees on their undercanopy environments in high- and low-rainfall savannas. Journal of Applied Ecology 30: 143–155.
- 28. Grouzis M, Akpo L-E (1997) Influence of tree cover on herbaceous above- and below-ground phytomass in the Sahelian zone of Senegal. Journal of Arid Environments 35: 285–296.
- 29. Schade JD, Sponseller R, Collins SL, Stiles A (2003) The influence of Prosopsis canopies on understorey vegetation: Effects of landscape position. Journal of Vegetation Science 14: 743–750.
- 30. Weltzin JF, Coughenour MB (1990) Savanna tree influence on understorey vegetation and soil nutrients in northwestern Kenya. Journal of Vegetation Science 1: 325–334.
- 31. Dohn J, Dembélé F, Karembé M, Moustakas A, Amévor KA, et al. (in press) Tree effects on grass growth in savannas: Competition, facilitation and the stress-gradient hypothesis. Journal of Ecology DOI: 10.1111/1365-2745.12010.
- 32. Durr PA, Rangel J (2002) Enhanced forage production under Samanea saman in a subhumid tropical grassland. Agroforestry Systems 54: 99–102.
- 33. Stuart-Hill GC, Tainton NM (1989) The competitive interaction between Acacia karroo and the herbaceous layer and how this is influenced by defoliation. Journal of Applied Ecology 26: 285–298.
- 34. Mordelet P, Menaut JC (1995) Influence of trees on above-ground production dynamics of grasses in a humid savanna. Journal of Vegetation Science 6: 223–228.
- 35. Ludwig F, de Kroon H, Prins HHT, Berendse F (2001) Effects of nutrients and shade on tree-grass interactions in an East African savanna. Journal of Vegetation Science 12: 579–588.
- 36. Xu C, Zhang M, Liu M, An SQ, Sheng S (2012) Interspecific effects on plant size inequality: evidence from a temperate savanna community. Plant Ecology 213: 225–235.
- 37. Ludwig F, Dawson TE, de Kroon H, Berendse F, Prins HHT (2003) Hydraulic lift in Acacia tortilis trees on an East African savanna. Oecologia 134: 293–300.
- 38. Ludwig F, Dawson TE, Prins HHT, Berendse F, de Kroon H (2004a) Below-ground competition between trees and grasses may overwhelm the facilitative effects of hydraulic lift. Ecology Letters 7: 623–631.
- 39. Dickie IA, Schnitzer SA, Reich PB, Hobbie SE (2005) Spatially disjunct effects of co-occurring competition and facilitation. Ecology Letters 8: 1191–1200.
- 40. Frost P, Medina E, Menaut J-C, Solbrig O, Swift M, et al. (1986) Responses of savannas to stress and disturbance. Biology International, Special Issue 10: 1–82.
- 41. Treydte AC, Looringh van Beeck FA, Ludwig F (2008) Heitkoenig IMA (2008) Improved quality of beneath-canopy grass in South African savannas: Local and seasonal variation. Journal of Vegetation Science 19: 663–670.
- 42. Govender N, Trollope WSW, Van Wilgen BW (2006) The effect of fire season, fire frequency, rainfall and management on fire intensity in savanna vegetation in South Africa. Journal of Applied Ecology 43: 748–758.
- 43. Higgins SI, Bond WJ, February EC, Bronn A, Euston-Brown DIW, et al. (2007) Effects of four decades of fire manipulation on woody vegetation structure in savanna. Ecology 88: 1119–1125.
- 44. Bransby DI, Tainton NM (1977) The disc pasture meter: possible applications in grazing management. Proceedings Grassland Society of South Africa 5: 115–118.
- 45. Trollope WSW, Potgieter ALF (1986) Estimating grass fuel loads with a disc pasture meter in the Kruger National Park. Journal of the Grassland Society of Southern Africa 3: 148–151.
- 46. Walker BH, Stone L, Henderson L, Vernede M (1986) Size structure analysis of the dominant trees in a South African savanna. South Africa Journal of Botany 52: 397–402.
- 47. Pinheiro JC, Bates DM (2004). Mixed-effects modes in S and S-Plus. Springer, New York.
- 48. Burnham KP, Anderson DR (2002) Model selection and multimodel inference: a practical information-theoretic approach. 2nd Edition. Springer.
- 49. R Development Core Team (2012). R: A language and environment for statistical computing. R Foundation for Statistical Computing, Vienna, Austria. ISBN 3-900051-07-0, URL http://www.R-project.org.
- 50. Burrows WH, Carter JO, Scanlan JC, Anderson ER (1990) Management of savannas for livestock production in north-east Australia: contrasts across the tree-grass continuum. Journal of Biogeography 17: 503–512.
- 51. Bertness MD, Callaway R (1994) Positive interactions in communities. Trends in Ecology & Evolution 9: 191–193.
- 52. Brooker RW, Callaghan TV (1998) The balance between positive and negative plant interactions and its relationship to environmental gradients: a model. Oikos 81: 196–207.
- 53. Callaway RM, Brooker RW, Choler P, Kikvidze Z, Lortie CJ, et al. (2002) Positive interactions among alpine plants increase with stress. Nature 417: 844–848.
- 54. Gómez-Aparicio L, Zamora R, Gómez JM, Hódar JA, Castro J, et al. (2004) Applying plant facilitation to forest restoration: A meta-analysis of the use of shrubs as nurse plants. Ecological Applications 14: 1128–1138.
- 55. Riginos C, Grace JB, Augustine DJ, Young TP (2009) Local versus landscape-scale effects of savanna trees on grasses. Journal of Ecology 97: 1337–1345.
- 56. Hoffmann WA, da Silva ER Jr, Machado GC, Bucci SJ, Scholz FG, et al. (2005) Seasonal leaf dynamics across a tree density gradient in a Brazilian savanna. Oecologia 145: 307–316.
- 57. Ratnam J, Bond WJ, Fensham RJ, Hoffmann WA, Archibald S, et al. (2011) When is a ‘forest’ a savanna and why does it matter? Global Ecology & Biogeography 20: 653–660.
- 58. Ludwig F, de Kroon H, Berendse F, Prins HHT (2004b) The influence of savanna trees on nutrient, water and light availability and the understorey vegetation. Plant Ecology 170: 93–105.
- 59. Mlambo D, Nyathi P, Mapaure I (2005) Influence of Colophospermum mopane on surface soil properties and understorey vegetation in a southern African savanna. Forest Ecology & Management 212: 394–404.
- 60. Treydte AC, Riginos C, Jeltsch F (2010) Enhanced use of beneath-canopy vegetation by grazing ungulates in African savannahs. Journal of Arid Environments 74: 1597–1603.
- 61. du Toit J, Biggs H, Rogers K (2003) The Kruger experience: ecology and management of savanna heterogeneity. Island Press
- 62. Treydte AC, Heitkonig IMA, Prins HHT (2007) Ludwig F (2007) Trees improve grass quality for herbivores in African savannas. Perspectives in Plant Ecology Evolution & Systematics 4: 197–205.
- 63. Milchunas DG, Lauenroth WK, Chapman PL, Kazempour MK (1989) Effects of grazing, topography, and precipitation on the structure of a semiarid grassland. Vegetatio 1: 11–23.
- 64. Callaway RM, Kikodze D, Chiboshvili M, Khetsuriani L (2005) Unpalatable plants protect neighbors from grazing and increase plant community diversity. Ecology 86: 1856–1862.
- 65. Kikvidze Z, Khetsuriani L, Kikodze D, Callaway RM (2006) Seasonal shifts in competition and facilitation in subalpine plant communities of the central Caucasus. Journal of Vegetation Science 17: 77–82.