Limiting factors for the short-term recruitment of savanna trees at woodland islands in the Western Soutpansberg, South Africa

List of contents

Abstract

Zusammenfassug

I. Introduction
1.1 Theoretical background
1.1.1 Classification of savanna types
1.1.2 Termites
1.1.3 Competition and facilitation
1.1.4 Fire
1.1.5 Browsing and seed predation
1.1.6 Life history traits
1.2 Research questions

II. Methods
2.1 Site description: Lajuma and the Western Soutpansberg
2.2 Selection of study areas
2.3 Tree species
2.4 Seed collection and seed weight classes
2.5 Experimental set-up
2.6 Giving-up density
2.7 Soil sampling
2.8 Soil temperature measurements
2.9 Soil analysis
2.10 Vegetation mapping
2.11 Data evaluation

III. Results
3.1 Seed weight in Mimusops zeyheri and Syzygium legatii seeds
3.2 The influence of termite-affected soil on germination and seedling
establishment
3.3 Field experiment II: Browsing and competition
3.4 Plant pot experiment II: The influence of shading
3.5 Giving-up density
3.6 Soil type, soil temperature, soil moisture content and bulk density
3.7 Soil chemistry
3.8 Vegetation mapping

IV. Discussion
4.1 Limitations of the study
4.2 The influence of seed weight on short-term tree recruitment
4.3 The influence of soil characteristics on short-term tree recruitment
4.4 The influence of light intensity on short-term tree recruitment
4.5 The influence of competition with grasses on short-term tree recruitment
4.6 The influence of seed predation and browsing on short-term tree recruitment
4.7 Limiting factors for the tree recruitment at the study area

V. Conclusions

VI. Outlook

VII. Acknowledgements

VIII. Literature

Appendix

List of Figures

2.1 Map of the Limpopo Province, South Africa.

2.2 Aerial photograph of the patches taken in 2003 with marked experimental patches

2.3.1 Mimusops zeyheri

2.3.2 Syzygium legatii

2.5.1 Transport of the cages to the patches

2.5.2 Experimental set-up for the field experiment I

2.5.3 Nursery

2.5.4 Plot excluding competition and browsing

2.5.5 Experimental set-up for the field experiment on browsing and competition

2.6 Experimental set-up for the GUD

2.7 Sampling sites per patch

2.10.1 Mapping sites per patch

2.10.2 Patch border

2.10.3 Browsing intensity for the example of Hyperacanthus amoenus

3.1.1 Distribution of seed weight in Mimusops zeyheri

3.1.2 Distribution of seed weight in Syzygium legatii

3.5 GUD: Statistical differences in seed predation by rodents for distinguished habitat types of tree patches in the Western Soutpansberg, South Africa

3.6 Soil moisture content, bulk density and soil temperature in different habitat types of a tree patch in relation to patch quarter exposition

3.7.1 Differences in nutrient contents between habitat types of tree patches in the Western Soutpansberg, South Africa

3.7.2 Variance of pH values among different habitat types of tree patches investigated in the Western Soutpansberg, South Africa

3.8.1 Seedling number, light intensity, competition and browsing measured for Gymnosporia seedlings in different habitats

3.8.2 Seedling number, light intensity, competition and browsing measured for Tarenna seedlings in different habitats

3.8.3 Seedling number, light intensity, competition and browsing measured for Hyperacanthus seedlings in different habitats

3.8.4. Seedling number, light intensity, competition and browsing measured for

Canthium seedlings in different habitats

3.8.5 Seedling number, light intensity, competition and browsing measured for Dovyalis seedlings in different habitats

3.8.6 Seedling number, light intensity, competition and browsing measured for Maytenus seedlings in different habitats

3.8.7 Seedling number, light intensity, competition and browsing measured for Psydrax seedlings in different habitats

3.8.8. Seedling number, light intensity, competition and browsing measured for Eugenia seedlings in different habitats

3.8.9 Total seedling number in different habitat types of a tree patch in relation to patch quarter exposition

3.8.10 Light intensity in different habitat types of a tree patch in relation to patch quarter exposition

List of Tables

2.2 North-South and East-West diameter of selected patches and their dominant tree species

2.4 Weight classes determined for Syzygium and Mimusops

2.5 Sowing dates and number of seeds per species for the germination experiments

3.2.1 Seedling numbers for Syzygium legatii and Mimusops zeyheri per weight class and soil sampled from increasing distances to a termite mound

3.4.1 Germination rates for shaded and non-shaded seeds of Syzygium legatii and Mimusops zeyheri

3.5 Results from Kruskal-Wallis Tests analysing for differences in seed predation in three habitat types of tree patches in the Western Soutpansberg, South Africa for twelve sampling days

3.6.1 General Linear Models testing for differences in soil moisture content (smc), bulk density (bd) and soil temperature (st) between 96 plots differing in exposition, Habitat type and location investigated in the Western Soutpansberg, South Africa

3.7.1 General Linear Models testing for differences in the carbon and nutrient contents between 96 plots differing in exposition, habitat type and location investigated in the Western Soutpansberg, South Africa

3.7.2 General Linear Models testing for differences in the pH values between 96 plots differing in exposition, habitat type and location investigated in the Western Soutpansberg, South Africa

3.7.3 Results of partial correlations between different nutrient concentrations (nc) across the 12 patch areas controlling for the effects of habitat type

3.8.1 Complete species list of tree seedlings and saplings found at the studied patches.

3.8.2 General Linear Models testing for differences in the seedling distribution between 96 plots differing in exposition, habitat type and location investigated in the Western Soutpansberg, South Africa

3.8.3 General Linear Models testing for differences in light intensity between 96 plots differing in exposition, habitat type and location investigated in the Western Soutpansberg, South Africa

Abstract

A key question of plant ecology is which factors control the local distribution of plant species and plant communities. Thus the appearance and persistence of scattered tree‑dominated fertile patches in generally nutrient-poor savanna grasslands is an interesting ecological phenomenon. The tree islands alter structural and spatial variability of the environment and thus are very important for floral and faunal biodiversity. Where conditions are favourable such patches are known to increase in size until they merge with each other and a closed canopy forest builds up. However, in dry areas successive invasion into grassland is blocked and there is little or no spread outwards. Mosaic shifts do not occur. Tree recruitment may be limited by moisture availability, soil characteristics or other habitat features like fire, competition and herbivory. On the other hand, trees as well as termites are known to enhance soil moisture and nutrient availability thus facilitating recruitment.

Research for the presented study was conducted from September 2005 to January 2006 in the Western Soutpansberg Mountain Range, South Africa. In two series of plant pot experiments effects of termite-affected soil, canopy shade and seed size on germination and seedling recruitment have been investigated. In the first case seeds were sown in soil sampled from three increasing distances from a termite mound. In the second experiment seedlings were raised either under full sunlight or imitated canopy shade. Furthermore, three field experiments were conducted investigating seed predation, response of seedlings to soil characteristics in increasing distances to termite mounds and combined effects of herbivory and competition. Additionally, seedling recruitment was studied at eight tree islands and soil samples from 96 plots varying in exposition, habitat type and location of sampling site were analysed.

Results revealed that tree recruitment was influenced by complex interactions between plant facilitation and competition, herbivory and abiotic environmental stress. Soil moisture availability and competition with grasses seem to be the primary factors limiting a rapid expansion of the tree islands into the surrounding grassland. Also the relationship between the prevailing wind direction of mist-loaded winds and moisture distribution influences chances for recruitment. Seed predation compared to the patch centre increased at the patch border and even more so in the open grassland, but the relative high presence of tree seedlings especially at the patch border suggests that seed availability is not limiting recruitment. Seedlings from larger seeds had a higher chance of survival and establishment. In once established saplings additionally to soil moisture and competition from grasses, browsing was found to play a major role in limiting tree recruitment.

Zusammenfassung

Welche Faktoren die örtliche Verteilung von Pflanzenarten bestimmen ist eines der Schlüsselthemen der Pflanzenökologie. Daher ist das Auftreten und Fortbestehen von verstreuten baumdominierten Fertilitätsinseln in nährstoffarmen Savannen ein ökologisch interessantes Phänomen. Waldinseln erhöhen die strukturelle und räumliche Variabilität der Umwelt und sind daher von beträchtlicher Bedeutung für die Biodiversität. Unter günstigen Umständen können solche Inseln sich im Laufe der Zeit ausdehnen und zu geschlossenen Wäldern verschmelzen. In trockenen Gebieten jedoch wird die Sukzession behindert und sie dehnen sich nur wenig oder nicht aus. Mosaikzyklen sind hier nicht zu beobachten. Baumverjüngung könnte durch Wasserverfügbarkeit, Bodenbeschaffenheit oder andere Habitateigenschaften wie Feuer, Konkurrenz und Herbivorie limitiert sein. Andererseits ist bekannt, dass sowohl Bäume als auch Termiten Wasser- und Nährstoffverfügbarkeit in Savannen lokal entscheidend verbessern und die Verjüngung dadurch begünstigen.

Die Untersuchungen, die der vorliegenden Studie zugrunde liegen, wurden zwischen September 2005 und Januar 2006 im Western Soutpansberg, Südafrika durchgeführt. In zwei Serien von Blumentopf-Experimenten wurde die Bedeutung von durch Termiten beeinflussten Böden, Beschattung und Samengröße auf Keimung und Keimlingsentwicklung untersucht. Im ersten Fall wurden die Baumsamen in Boden, der in wachsenden Entfernungen zu einem Termitenhügel entnommen wurde, gepflanzt. Im zweiten Fall wurden Keimung und Keimlingsentwicklung unter voller Sonneneinstrahlung und imitiertem Baumschatten beobachtet. Weiterhin wurden drei Feldexperimente durchgeführt. Untersucht wurden Samenprädation, Reaktion von Keimlingen auf mit wachsender Distanz zu Termitenhügeln variierende Bodenbedingungen und Einfluss von Wildverbiss und Konkurrenz. Zusätzlich wurde die Baumverjüngung in acht Waldinseln studiert und 12 Bodenproben pro Bauminseln, die in Exposition und Habitattyp der Probefläche variierten, wurden analysiert.

Die Ergebnisse zeigen, dass die Baumverjüngung durch komplexe Interaktionen von Konkurrenz und Begünstigung zwischen Pflanzen, Herbivorie und abiotischem Stress beeinflusst wird. Bodenwasserverfügbarkeit und Konkurrenz mit Gräsern scheinen dabei die primären Faktoren zu sein, die eine rasche Expansion der Waldinseln verhindern. Auch die Beziehung zwischen vorherrschender Windrichtung bei Nebel und Feuchtigkeitsverteilung beeinflusst die Chancen für Baumverjüngung. Die Samenprädation steigt im Vergleich zum Inneren der Waldinsel an ihrer Grenze leicht und im umliegenden Grasland stark an, jedoch deutet das relativ häufige Vorkommen von Keimlingen besonders an den Inselgrenzen darauf hin, dass die Samenverfügbarkeit nicht limitierend für die Verjüngung ist. Keimlinge größerer Samen haben höhere Etablierungs- und Überlebenschancen. Das Heranwachsen etablierter Jungbäume wird neben Wasserverfügbarkeit und Konkurrenz entscheidend durch Wildverbiss erschwert.

I. Introduction

1.1 Theoretical background

1.1.1 Classification of savanna types

Savanna ecosystems are characterised by the co-dominance of two different life forms: grasses and trees. Although the relative representation of these life forms varies considerably across savanna types, they typically comprise communities with a continuous herbaceous layer and a discontinuous stratum of shrubs and trees (Solbrig et al. 1996, Higgins et al. 2000, Sankaran et al. 2004). The floristic composition of the different savannas is quite variable. In the herbaceous layer, two families dominate throughout all the savannas: the Gramineae and Cyperaceae. In the woody layer there is no such uniformity and a different mix of species, genera and families prevails in each continent, and within continents in each region, due to differences in the physical environment and human activity (Solbrig et al. 1996).

As savannas occupy one eighth of the global land surface and are home to a large and steadily growing proportion of the world’s human population, its rangeland and livestock, they are of increasing importance to human welfare and economy. Despite this fact the origin, nature and dynamics of savannas are poorly understood (Scholes & Archer 1997, Sankaran et al. 2004). An operational understanding of tree-grass coexistence is essential for understanding savanna function and for predicting its response to future environmental change (Higgins et al. 2000).

The main controls on tree recruitment are from fire, rainfall, herbivory and competition with grasses (Mistry 2000). Recruitment furthermore is influenced by soil fertility, geomorphology and topography (Huntley & Walker 1982, Solbrig et al. 1996, Higgins et al. 2000). These determinants are predicted to operate at all ecological scales from landscapes to local patches, but their relative importance differs with scale (Solbrig et al. 1996). Explanations for the coexistence of grasses and trees so far either concentrated on competition-based mechanisms, where niche separation with respect to limiting resources such as water leads to tree-grass coexistence, or demographic mechanisms, where factors like fire, herbivory and rainfall availability promote tree‑grass persistence through their dissimilar effects on different life-history stages of trees (Sankaran et al. 2004). Understanding the variation in tree recruitment needs not only an understanding of how and to which intensity the above mentioned factors influence tree recruitment but also an understanding of the life history of savanna trees (De Steven 1991b, Moles & Westoby 2004).

Southern African savannas can be divided into two main types: sweetveld and sourveld. Sweetveld occurs in lowland, warm, arid areas on fertile soils, whereas sourveld occurs in highland, cool, wet infertile areas (Scholes 1991, Mistry 2000). The soils of arid eutrophic (sweetveld) savannas are characterised by higher ion exchange capacity and a higher accumulation of bases than those of moist distrophic (sourveld) savannas. They contain more humus and the influences of the base rock and relief are more obvious than in the moist savannas. The productivity of the vegetation is limited by rainfall rather than by nutrient content as in the moist savannas. Here soils are – as a result of deep reaching weather-beatening of the base rock, higher decomposition rates and progressive washing-out – nutrient poor and contain only little humus. If the annual rainfall exceeds 500-600 mm washing-out processes prevail accumulation, and most soil types show an acid reaction as a result of an impoverishment in exchangeable nutrients. Free carbonate is missing (Schultz 2002).

According to Scholes (1991) at a generalisation level, the eutrophic/distrophic distinction appears to be valid and useful to classify entire regions as fertile or infertile, but it tends to hide the fact that within a landscape both savanna types may consist of mosaics of fertile and infertile sites. The origin of fertility differences can be geological, geomorphologic, anthropogenic, or biotic. There are several processes which result in the creation and maintenance of nutrient-rich and nutrient-poor patches within the landscape, including patterns of herbivory and nutrient cycling through trees and termite mounds (Scholes 1991). Savanna trees affect herbaceous phenology, production, and biomass allocation as well as species composition (Scholes & Archer 1997). Also the theory of shifting landscape mosaics resulting from disturbances like fire, herbivory and termite activity is often discussed (Callaway & Davis 1993, Fuhlendorf & Engle 2004). The formation of localised nutrient enriched patches is suggested to be of great importance to the function of the savanna as a whole, particularly in a generally nutrient-poor savanna environment such as investigated in this study (Scholes 1991).

1.1.2 Termites

In the seasonally dry savannas of the tropics many microorganisms are inactive for the major part of the year because their activity is determined by soil moisture regimes. This gives an important role to termites, the dominant group of soil animals in these ecosystems. These have a significant effect on the nutrient cycling and vegetation of local patches (Holt & Coventry 1991, Mando 1997), because their activity enhances decomposition and hence nutrient release in the soil and leads to a localised accumulation of bases (Mando 1997). Part of this enrichment results from fine mineral particles brought up from the subsoil for nest construction (Mistry 2000). The increased clay content typical of termite mounds leads to a greater ion exchange capacity, which helps to retain nutrients (Scholes 1991). Termite foraging and mound-building also causes changes in soil structure (Lee and Wood 1971, Scholes 1991, Mando 1997, Mistry 2000). Throughout the soil profile, macropores with irregular shapes and with different diameter sizes are created. Through these changes in soil structure water infiltration, water availability and drainage are greatly improved and the runoff volume after rain decreases (Mando 1997). Lower bulk density of termite-affected soil compared with unaffected soil may result from increased porosity, due to termite galleries and incorporation of organic matter in the soil. Furthermore fungus-growing termites carry fungi as spores and hyphae on their body, which fall down on fresh substrata while termites move through the soil. They thus serve the spread of fungi, which additionally to termites occupy an important role in breaking down cellulose. Relative to the surrounding soil termite mounds have increased pH values and contain more carbon, nitrogen and more exchangeable calcium, magnesium and potassium. Redistribution of these elements by erosion and decomposition enriches soils in the vicinity of termites.

Compared to the surroundings, higher plant species diversity and biomass production have been measured close to termite mounds (Lee & Wood 1971, Bourliere 1983, Mando 1997, Mistry 2000). In particular, tree density is known to increase in the vicinity of old termite mounds (Glover et al. 1964, Yeaton 1988). Lee & Wood (1971) reported that protection from fires and from waterlogging as well as higher nutrient levels are the main factors favouring development of woodland on large termite mounds. Mounds range in size from structures only a few centimetres tall to the colossal mounds built by some species of Macrotermitinae which reach 9 m in height and 25 m in diameter (Lee & Wood 1971). Despite the great functional importance of savanna insects as ecosystem engineers, they have largely been ignored by savanna ecologists (Holt & Coventry 1991).

1.1.3 Competition and facilitation

For woody plants with long life-spans and low post-establishment mortality rates, seedling recruitment is a critical life-history stage (De Steven 1991a, Scholes & Archer 1997). Grasses may regulate tree recruitment directly through competition for light, water and nutrients or indirectly through influencing fire frequency and intensity. It is evident that grasses may have a strong negative effect on young trees which are within the flame zone of grass-layer fires. But once trees overtop the herbaceous vegetation they are better competitors for light, while grasses are better competitors for nutrients and water. Although grasses reduce the emergence, growth and survival of woody seedlings the competitive reduction is usually not strong enough to result in high mortality rates or complete exclusion of trees in the absence of herbivores and fires (Scholes & Archer 1997). It is supposed that tree recruitment occurs in periods of high moisture availability, when competition from grasses is minimal. Trees quickly overtop the grass layer if grass growth, fire and browsing are limited, but the regeneration niches of savanna tree species are not well known. In patches of closed forest recruitment is often controlled by gaps in the canopy generated by disturbances such as windstorms and lightning (Clarke & Kerrigan 2002).

The presence of woody plants can alter the composition, spatial distribution and productivity of grasses in savannas (Joffre & Rambal 1993). The effect of trees on grasses ranges from positive to neutral to negative (De Steven 1991b). At the scale of the tree, species composition of the grass layer may change along gradients extending from the bole to the canopy drip line and into the adjoining inter-tree zone. Increased herbaceous production beneath tree canopies was associated with lower soil temperatures, lower plant water stress, greater soil organic matter concentrations and microbial biomass compared to those of adjacent grassland away from tree canopies. But soil properties and microclimate are changing with increasing duration of tree occupation. Under young and small trees facilitation could be more important than competition and grass production is enhanced; as trees become larger competition may become more important and reduce herbaceous production (Weltzin & Coughenour 1990, Scholes & Archer 1997). Negative effects may result from rainfall interception, litter accumulation, shading, root competition, or a combination of these factors; and the effect depends on the leaf area, canopy architecture, and rooting pattern of the tree (Borchert et al. 1989, Scholes & Archer 1997). Because of the potentially positive effects of young trees on grasses, herbaceous diversity and production can be greater in areas with low tree densities than where there are no trees, but the effect is reversed at high tree densities (Scholes & Archer 1997).

Woody plants often produce nitrogen-rich litter that decomposes more rapidly than grass litter in the ecosystems in which both occur. This can elevate available nitrogen levels under woody plants compared with parts of the grassland where woody plants are not present, and greater nitrogen availability tends to intensify above-ground competition for light and favours woody plants in competition with grasses (Siemann & Rogers 2003). As woody plant cover increases, grass productivity typically declines (Scholes & Archer 1997).

Tree-tree interactions influence the density and pattern of woody plant distribution across savanna landscapes and hence patterns of grass biomass and distribution. Regular savanna tree distribution suggests competition, while clumped tree distribution suggests facilitation. Clumped tree dispersal often occurs in response to variations in local topography such as termite mounds, fire patchiness and soil depth (Scholes & Archer 1997).

It is generally believed that the seedlings of many savanna species are shade intolerant and that high grass biomass can suppress tree seedlings. Other evidence suggests that some savanna tree seedlings are shade tolerant but cannot tolerate droughts during the wet growing season (Higgins et al. 2000). Li & Wilson (1998) showed that simultaneous shading, increasing water availability and fertilisation with nitrogen favoured woody seedlings in competition with grasses. Water availability under tree canopies usually is higher than in the grassland because of shading (Harrington 1991, Siemann & Rogers 2003). Light reductions under the canopies could be responsible for the unusually high success of tree seedlings that have been observed to recruit near established trees in grasslands (Siemann & Rogers 2003). Competition between trees may occur but is often counterbalanced by elements of facilitation, especially during the phase in which tree seedlings are established (Scholes & Archer 1997).

1.1.4 Fire

Although the effects of fire on tree recruitment could not be investigated in the presented study some information on its significance for tree recruitment in savanna ecosystems shall be given. Fire plays a critical role for tree recruitment and structure of plant communities and can have both constructive and destructive effects upon them (Tyler 1995, Freckleton 2004). After fire events a flush of seedling establishment was often observed, most probably as a result of temporary reductions in competition and herbivory as well as altered nutrient levels. In some savanna species fire events are even needed to break seed dormancy. Fire therefore sometimes has a positive effect on seed germination and seedling establishment (Tyler 1995) while on the other hand it prevents recruitment of seedlings and saplings into adult size class (Higgins et al. 2000). Seedlings and saplings within the flame zone of grass fires are usually destroyed. Only stems which are tall and thick enough are able to resprout (Scholes & Archer 1997, Higgis et al. 2000). Effects of fire also interact with other ecological factors like grazing and browsing (Freckleton 2004). Models of fire‑grazing interaction state that grazing and fire interact through a series of positive and negative feedbacks to cause a shifting mosaic of vegetation pattern across the landscape (Fuhlendorf & Engle 2004). High grass biomass can affect tree biomass by fuelling fires. Grazing reduces the fuel load and hence affects fire frequency, intensity, or continuity of spread. Browsing on the other hand helps to keep trees within the flame zone, while fires keep trees browsable (Scholes & Archer 1997).

1.1.5 Browsing and seed predation

Seed predation as well as browsing can have considerable negative effects on tree recruitment (Goldberg 1985, Ostfeld et al. 1997, Rooney et al. 2000). Rodents especially are known to have profound significance for seed availability and thus can cause strong reductions in recruitment of seedlings (Ostfeld & Canham 1993). As rodent activity usually varies with habitat types also reductions in seedling establishment differ between contrasting habitats (Hulme 1994). Browsing often results in significant damage to woody plants and may inhibit recruitment of saplings into adult stage (Higgins et al. 2000, Rooney et al. 2000). Similar to seed predation browsing may also vary with habitat type and also with tree species as some species are less palatable than others due to chemical or mechanical defences (Rooke et al. 2004). As savanna trees alter structural complexity of the environment and offer shade and shelter several animals are attracted to trees. Perching birds and herbivores that take refuge in the shade of trees enhance the local nutrient cycle by their faeces and may deposit seeds of other trees and shrubs whose germination and establishment may be favoured in the subcanopy environment (Scholes & Archer 1997). Scholes (1991) reported that mammalian herbivores have been observed to spend up to 6-fold more time on the nutrient rich patches than would be expected from a random distribution model while productivity on these sites was only twice that in surrounding nutrient-poor areas. He therefore suggested that it is likely that defecation exceeds consumption on the sites and the nutrient flux due to herbivory is inwards, not outwards.

1.1.6 Life history traits

Seed size is a relevant life history trait in many tree species. Many hitherto existing studies showed that large-seeded species have higher rates of establishment due to being more resistant to many hazards including drought (Leishman & Westoby 1994, Jurado & Westoby 1992), shading (Sork 1987, Saverimuttu & Westoby 1996, Grime & Jeffrey 1965), nutrient shortage (Stock et al. 1990) and competition from other seedlings (Harper et al. 1970, Battaglia et al. 2000) and established vegetation (Reader 1993), thus having higher rates of survival during seedling establishment (Moles & Westoby 2004, Leishman et al. 1995). Light seeds on the other hand are usually produced in greater numbers which gives small‑seeded species an initial advantage over large-seeded ones (Moles & Westoby 2004). Rodents, due to the usually increasing energy gain, prefer large seeds over small ones (Sork 1987, de Steven 1991a, Chambers & MacMahon 1994, Hulme 1996). Although large seeds generally appear to be advantageous in seedling establishment, this may be a considerable negative effect influencing seed fate.

In general large-seeded species provide a higher metabolic reserve, therefore producing larger seedlings (Stock et al. 1990, Leishman & Westoby 1994a). Some previous evidence also suggests that larger seeds tend to have greater percentage seedling emergence (Stanton 1984, Wulff 1986a, Winn 1988, Dalling & Hubbel 2002, but see Stock et al. 1990, Moles & Westoby 2004) and a more extensive root system conferring on them a greater drought tolerance (Leishman & Westoby 1994, Hoffmann 1996). Within species the optimal seed weight is that which maximizes seedling fitness per unit investment. Seeds smaller than the optimum have low fitness while in seeds larger than the optimum a waste of resources occurs that could otherwise be used to produce higher seed numbers. Such ability to establish in different habitat types is supposed to be different for small‑seeded and large-seeded species and also within species between smaller and larger seeds (Vaughton & Ramsey 1998, Wulff 1986a,b).

1.2 Research questions

The aim of the investigation was to evaluate the impacts of seed size, soil fertility modified by trees and termites, light, competition and herbivory on short-term tree recruitment of termite mound tree islands. To gain information about factors influencing the persistence and stability in size of patches occurring on a grassland terrace in the Western Soutpansberg Range, South Africa two series of plant pot and three field experiments were conducted. Additionally recruitment was studied by seedling and sapling mapping at eight tree islands. The term of short‑term recruitment in this work has been defined as the time from germination to sapling stage. Because of the time limit of the research the growing period for seedlings in the experiments conducted was restricted to approximately twelve weeks. The effects of rainfall and fire could therefore not be tested.

The creation and long-term persistence of fertile patches within infertile savannas is a widespread phenomenon (Scholes 1991). As mentioned above the foraging and mound‑building activities of termites lead to physical and chemical changes in the soil structure in the vicinity of the mound (Scholes 1991), and the erosional redistribution of termite mound materials can supply infertile surface soil with a nutrient enriched soil amendment (Holt & Coventry 1991) which has an important influence on the nutrient cycling of soils and on their water balance. The first trees establishing in grassland may indirectly favour their seedlings in competition with neighbouring herbaceous vegetation by increasing available nutrients, water infiltration and soil organic matter with nitrogen-rich litter and by reducing light levels under their canopies (Siemann & Rogers 2003, Belsky 1994, Solbrig at al. 1996, Tolsma et al. 1987). Woody species can create microhabitats in which species other than those dominating in the grassland are able to establish and hence trees are of considerable importance for the biodiversity of savanna landscapes. Woody species producing large seeds may have higher seedling survival rates because they seem to be more resistant to hazards (Belsky 1994, Leishman & Westoby 1994). Furthermore, several animals are attracted to trees which offer them forage, shade and shelter (Belsky 1994) and hence influence the nutrient cycling and grass-tree coexistence in savannas.

On the basis of these considerations the following questions will be examined:

- Does seed weight influence germination success and seedling performance?
- How do soil characteristics vary with distance to the centre of tree islands, and do these variations in soil characteristics influence germination rate and short-term recruitment of trees? Do fertility differences result in profound differences in the species composition?
- In which way does light availability influence seed germination and the short-term recruitment of trees?
- Does competition with grasses reduce the germination and survival of seedlings?
- Does seed predation and browsing limit short-term recruitment of trees?

II. Methods

2.1 Site description: Lajuma and the Western Soutpansberg

illustration not visible in this excerpt

Figure 2.1: Map of the Limpopo Province, South Africa. The arrow points out the investigation area in the Soutpansberg Mountain Range. Map was changed after www.sa venues.com/maps/limpopo-relief.htm 03.05.06.

Selected for investigation was a grassland terrace in the upper reaches of the Soutpansberg Mountain Range, South Africa belonging to the Lajuma Reserve. The location of Lajuma is shown in Figure 2.1. Generally the Soutpansberg consists of sourveld, but vegetation varies from montane grassland to woodland, thicket and groundwater forest. Due to its location and a diversity of soil and climatic zones, the area contains a remarkable diversity of plants and animals. More than 2.500 vascular plant taxa are known to occur in the Soutpansberg mountain range of which close to 600 are trees (Gaigher, pers. comm.). The climate of the Soutpansberg area is strongly influenced by the east-west orientated mountain range that represents an effective barrier between the south-easterly maritime climate of the Indian Ocean and the continental climate influences coming from the north. Wind patterns play an important role in determining local climates. In terms of seasonality the Soutpansberg region has two clearly differentiated seasons: the cool dry season from May to August and the warm, wet season from October to March (www.soutpansberg.com 07.02.06). Topography is rugged with deep valleys and high cliffs. The interplay between topography and macro-climate determines a mosaic of different habitats and micro-climates and, as yet, mostly undescribed assemblages of plant and animal communities. The Soutpansberg region is rather rural, traditional in appearance and lacking community development projects. Because of the subhumid to semi-arid climate the area is highly vulnerable to human degradation. It is supposed that about 65 per cent of the area is transformed by human activity. About 12 per cent are formally conserved. Population pressure, clearing for agriculture, forestry plantations and overgrazing have caused decline of extensive natural and semi-natural habitats (www.soutpansberg.com 07.02.06).

Lajuma, which is part of the Soutpansberg conservancy and of the Thavha Ya Muno Private Nature Reserve (about 5.000 ha), is situated in the Limpopo province 25 km east of Vivo and 45 km west of Makhado high up in the Soutpansberg mountain range. The reserve has a surface area of 430 ha (Gaigher, pers. comm.).

The territory selected for investigation can be typified by an interesting mosaic vegetation type consisting of short sour grassland on an infertile sandy substrate with scattered fertile patches of closed woodland. As can be seen from aerial photographs taken in 10‑year intervals the patches have been fairly stable in size over the last 50 years indicating a stable mosaic. Each island covers an area of on average 700 m². The sandy, well-drained soil is referred to as Hutton and originated from weathered sandstone and quartzite (Gaigher, pers. comm.). The dominant grass species between the islands is Loudetia simplex growing on nutrient-poor soils. Grass diversity increases around the patches. Each patch contains at least one termite mound, and the occurrence and plant species composition of the patches is believed to be influenced by various factors, especially moisture and nutrient availability, fire, competition and the impact of herbivores.

Average annual rainfall amounts to 730 mm, but varies considerably from year to year. Further precipitation results from relative frequent mist incidents connected to a prevailing south-eastern wind direction. The study area is inhabited by a variety of larger mammal species such as common duiker (Sylvicapra grimmia), bushbuck (Tragelaphus scriptus), kudu (Tragelaphus strepsiceros), klipspringer (Oreotragus oreotragus), bushpig (Patamochoerus larvatus), warthog (Phacochoerus aetiopicus), chacma baboon (Papio ursinus), vervet monkey (Cercopithecus aetiops) and a variety of birds, rodents and predators all of which play some role in the dynamics of the system. The termites occurring at the patches are Macrotermes natalensis and Odontotermes spp. (Gaigher, pers. comm.). Macrotermes as well as Odontotermes spp. belong to the order of Isoptera, the family of Termitidae and the subfamily of fungus-growing Macrotermitinae (Krishna & Weesner 1969).

2.2 Selection of study areas

illustration not visible in this excerpt

Figure 2.2: Aerial photograph of the patches taken in 2003 with marked experimental patches (WB = occurrence of Mountain Waterberry, MW = occurrence of Transvaal Red Milkwood)

The criteria for selecting the study patches have been patch size and the occurrence of either Transvaal Red Milkwood (Mimusops zeyheri) or Mountain Waterberry (Syzygium legatii). It was assumed that chosen patches have comparable conditions for tree recruitment. The selected patches are shown in Figure 2.2. Table 2.2 shows mean patch diameters and the dominant tree species of the selected patches.

Table 2.2: North-South and East-West diameter of the selected patches and their dominant tree species in the Western Soutpansberg, South Africa

illustration not visible in this excerpt

2.3 Tree species

The tree species chosen for the experimental study were Mountain Waterberry (Syzygium legatii, Figure 2.3.1) and Transvaal Red Milkwood (Mimusops zeyheri, Figure 2.3.2) as they were the only species with enough seeds available during the study period.

illustration not visible in this excerpt

Figure 2.3.1: Mimusops zeyheri

Transvaal Red Milkwood (Mimusops zeyheri, Sapotaceae) is a small to medium sized tree up to 15 m in height with a wide spreading crown. It appears at low altitudes in hot areas with adequate rainfall and is frequent on well wooded rocky hillsides, at the margins of evergreen forests and in dry open woodland and bushveld. Up to 4.5 cm long yellow fruits appear from April to October, normally carrying only one seed of about 2 cm in length inside (Palgrave 2002). Stock and game browse the leaves and young branches and eat the fruits from the ground (Joffe 2001).

illustration not visible in this excerpt

Figure 2.3.2: Syzygium legatii

Mountain Waterberry (Syzygium legatii, Myrtaceae) is a small to medium-sized tree which occurs in the mist-belt at forest margins and in scrub on rocky mountain slopes. It can be “common to dominant in the Soutpansberg.” (Palgrave 2002) The leaves are elliptic and leathery, 3.5-5.5 * 2-2.5 cm in size, and the under-surface is covered with very fine distinct net-veining. Flowers appear from December to July; the fruits are about 1 cm long and usually carry one seed (Palgrave 2002).

2.4 Seed collection, seed treatment and weight classes

The Syzygium seeds were collected directly from fresh reddish fruits lying under the Syzygium trees. The skin and flesh of the fruit was removed with a knife right up to the seed coat.

The Mimusops seeds were collected from baboon and vervet monkey dung. Following Joffe (2001) Mimusops was treated with fungicide before sowing. According to Carr (1994) germination varies from 20 to 55 percent and germination time ranges from 4 to 55 days.

The seed weights for both species were determined with electronic scales (UWE SC series, accuracy 0.1 g). Every seed was weighed individually. In Syygium seed weight varied much more than in Mimusops. Therefore larger weight classes had to be chosen here. For both species seven weight classes were determined as shown in Table 2.4.

Table 2.4: Weight classes determined for Syzygium and Mimusops

illustration not visible in this excerpt

2.5 Experimental set-up

Field experiment I: The influence of termite-affected soil on germination and seedling establishment

For the Syzygium recruitment experiments three similar sized patches (WB1 to WB3, Figure 2.2) were chosen to study seed germination and establishment. Around the main termite mound nine experimental plots (0.25 m²) were established (see Figure 2.5.2). In each plot 30 seeds were sown. Before sowing, groups of 30 seeds were weighed on an electronic balance to quantify the total seed weight for every experimental plot. In every plot the total seed weight was the same thus excluding potential effects of seed size on germination success as far as possible. The total weight for 30 seeds was 60.0 – 61.0 g. To prevent seed predation plots were covered with cages made of wood and sheeting material (Figure 2.5.1), which was buried approximately 10 cm into the soil. Additional shading resulting from cages was very low and it is supposed that the cages do not have considerable effect on soil temperature. Three plots were set up only 0.5 m away from the mound, three plots 3 m away and the last three 8 m away from the mound. A transect experiment would have been difficult as the patches usually contained more than one mound and distances would not be similar to all points of termite activity. In total 810 seeds from this species were needed for this experiment.

illustration not visible in this excerpt

Figure 2.5.1: Transport of the cages to the patches

For the Mimusops experiment five patches with the same criteria as mentioned above were chosen for the investigations (MW1-MW5, Figure 2.2). The set-up was similar to that of the Syzygium experiments, but only 20 seeds were planted per plot resulting in 900 planted seeds in total. The total seed weight per experimental plot was 9.0 g.

During the first four weeks after sowing the plots were watered with 4 mm/m² water three times a week, later the watering had to be stopped because of water shortage.

illustration not visible in this excerpt

Figure 2.5.2: Experimental set-up for the field experiment I

Plant pot experiment I: The influence of termite-affected soil on germination and seedling establishment

For the flower pot experiments 144 (3 * 48) seeds per tree species were grown on natural soil sampled from three different distances (0.5 m, 3 m, 8 m) away from an active termite mound with an sampling drill. Comparable to the field experiments this was done in order to examine the potential effects on germination rate and short-term recruitment caused by possible differences in nutrient availability and soil characteristics inside a tree patch. Possible confounding effects of water shortage were excluded by watering the plant pots every two to three days with approximately 8 mm water per 1 m² except on rainy days. To estimate the influence of termite-affected soil on seed germination, germination rate, biomass, leaf number and plant height were measured 14 weeks after sowing. Survival duration of seedlings after germination was also taken into account to examine the chance of seedling establishment. Before determining the biomass the plants were dried in an oven at 35 °C for ten hours. The plant pot experiments were implemented in a nursery (Figure 2.5.3) excluding animals by fencing.

illustration not visible in this excerpt

Figure 2.5.3: Nursery

Germination rate was calculated as follows:

Number of seedlings / 48 * 100 = germination rate (%)

Plant biomass was estimated by weighing the dry plant matter after drying each plant until it had a constant weight. The measurement of plant height was carried out with a measuring rod.

Field experiment II: Influences of browsing and competition on tree recruitment

In this experimental series the effects of different combinations of competition and browsing treatments are examined. These are:

[...]