Seasonal variation in soil seed bank size and species composition of selected habitat types in Maputaland , South Africa

Seasonal variation in seed bank size and species composition o f five selected habitat types within the Tembe Elephant Park. South Africa, was investigated. At three-month intervals, soil samples were randomly collected from five different habitat types: a, Licuati forest; b, Licuati thicket; c, a bare or sparsely vegetated zone surrounding the forest edge, referred to as the forest/grassland ecotone; d, grassland; and e, open woodland. Most species in the seed bank flora were either grasses, sedges, or forbs, with hardly any evidence o f woody species. The Licuati forest and thicket soils produced the lowest seed densities in all seasons. Licuati forest and grassland seed banks showed a two-fold seasonal variation in size, those o f the Licuati thicket and woodland a three-fold variation in size, whereas the forest/grassland ecotone maintained a relatively large seed bank all year round. The woodland seed bank had the highest species richness, whereas the Licuati forest and thicket soils were poor in species. Generally, it was found that the greatest correspondence in species composition was between the Licuati forest and thicket, as well as the forest/grassland ecotone and grassland seed bank floras.


INTRODUCTION
The Maputaland Centre of Plant Endemism (Van Wyk 1996) is known as a centre of high species diversity, rich in endemic plant and animal species (Scott-Shaw 1999;Van Wyk & Smith 2001) and is as such recog nized by the International Union for the Conservation of Nature (lUCN).A rare and unique vegetation type that is endemic to the Maputaland Centre of Plant Endemism is the Sand Forest (Licuati forest in Mozambique- Izidine et al. 2003) which houses a substantial number o f floristic endemics.However, the survival of many of Maputaland's endemic plant species is threatened by the rapid expansion of the human population and the associ ated demand for firewood, building material, medicinal plants, as well as land for agriculture and cattle grazing (Lawes el al. 2004;Kyle 2004).Even in formal con servation areas, elephants {Loxodonta africana) and fire could potentially threaten the long-term survival of many species.It is therefore o f the utmost importance that these biota-rich areas be conserved and managed properly.
To provide valuable information on the species com position o f Maputaland's diverse vegetation, several phytosociological studies have been done in recent years (e.g.Kirkwood & Midgley 1999;Matthews et al. 2001;Gaugris et al. 2004), but little is known about the veg etation dynamics and fianctional ecology.This is espe cially true for the Sand Forest vegetation where active management strategies are essential to ensure the long term survival of this vegetation type.Formulating successfiil strategies that will maintain the integrity of the Sand Forest, can only be achieved if they are based on a sound knowledge o f the vegetation dynamics and func tional ecology o f the system.
Seed bank dynamics are an important aspect o f the functional ecology of a vegetation type.Soil seed banks represent a pool of reproductive potential and a source of genetic inheritance and play an important role in vegeta tion establishment after a disturbance (Warr et al. 1993;Bakker et al. 1996).The absence of a soil seed bank has important consequences for the dynamics of a spe cies or vegetation type, because in such cases the veg etation will not be able to regenerate from a soil-stored seed bank after a disturbance.A soil seed bank, however, is not a static entity and the seed density and species composition of the soil seed bank flora constantly vary in space and time (Thompson & Grime 1979;Roberts 1986;Milberg & Hansson 1993;Crawford & Young 1998;De Villiers et al. 2004).
The present study is the first to investigate the soil seed banks in Maputaland.The main focus was on the Sand Forest and its associated vegetation types.The objectives of the study were to compare the seasonal changes in the soil seed bank in terms of a, seed density and b, species composition across five different habitats in the Tembe Elephant Park in northern KwaZulu-Natal.

THE STUDY AREA
The Tembe Elephant Park (26° 51.56' S-27° 03.25' S and 32° 24.17' E-32° 37.30' E) lies in the core of the Maputaland Centre of Plant Endemism, is 30 013 ha in extent and encloses extensive areas of pristine, endemic Sand Forest as well as other woodland, grassland and wetland vegetation types (Moll 1980;Matthews et al. 2001).
Sand Forest, also known as Licuati forest in Mozambique (Izidine et al. 2003), occurs under more arid conditions than other southern African forest types (Moll 1977;Moll & White 1980;Ward 1981;Low & Rebelo 1998;Kirkwood & Midgley 1999).It shows clear links to the tropical forest of southern Africa but has a relict character and as such is characterized by a low rate of regeneration with few seedlings and saplings (Von Maltitz et al. 2003).The Sand Forest can be sub divided into two related vegetation types: Licuati forest (also often called Tall Sand Forest) and Licuati thicket (also referred to as Short Sand Forest).The Licuati for est, with canopy heights up to 20 m, is characterized by diagnostic tree species such as Balanites maughamii, Cleistanthiis schlechteri, Dn'petes arguta, Newtonia hildebrandtii and Ptaeroxylum obliquum, and little under growth is present (Matthews et al. 2001;Gaugris et al. 2004).Licuati thicket is a near-impenetrable, short stature forest type with canopy height less than 10 m (Matthews et al. 2001;Izidine et al. 2003;Gaugris et al. 2004).Diagnostic species in the thicket vegetation are Croton pseudopulchellus, Hyperacanthus microphyllus, Psydrax fragrantissima and Ptaeroxylum obliquum.
The Licuati vegetation often occurs as a mosaic with a very specific type of grassland which is dominated by grass species such as Andropogon schirensis, Perotis patens, Pogonarthria squarrosa and the forb Tephrosia longipes subsp.longipes.The transition between the Licuati vegetation and the grassland is abrupt and is rep resented by a bare or sparsely vegetated zone bordering the forest edge, dominated by Perotis patens.This transi tional zone, which is only a few metres wide, is referred to as the forest/grassland ecotone in this study.
Several woodland types have been described for the Tembe Elephant Park ranging from sparse wood land, through open woodland to closed woodland.The open woodland, that was sampled for the seed bank study, is characterized by the geoxylic suffrutex Salacia kraussii, grasses such as Aristida stipitata subsp.stipitata, Panicum maximum, Pogonarthria squarrosa, and Themeda triandra and the tree species Acacia burkei, Afzelia quanzensis, Albizia adianthifolia, Garcinia livingstonei, Strychnos madagascariensis, S. spinosa and Terminalia sericea (Matthews et al. 2001;Gaugris et al. 2004).
The climate of Maputaland is characterized by hot, humid summers and cool, frost-free winters.According to weather data from the Sihangwane Weather Station in Tembe Elephant Park, the mean annual rainfall is 721 mm with a clear peak from October to April.The mean annual temperature for Tembe Elephant Park is 23.1°C, with absolute maximum and minimum temperatures of 45.0°C and 4.0°C (Gaugris et al. 2004).

METHODS
For the seed bank study, five different habitat types were selected in the southwestern part of the Tembe Elephant Park: a, Licuati forest; b, Licuati thicket; c, the forest/grassland ecotone; d, grassland; and e, woodland.
Sixty soil samples were collected within each of the five habitat types (5 x 60 samples) to determine the size and species composition of the soil seed bank.Soil sampling was carried out at three-month intervals for a period of 12 months.The top 100 mm of soil was col lected with a soil auger with a diameter of 57 mm because an investigation into the depth distribution of seed in the seed bank revealed that most seeds were found in this layer (Kellerman 2004).The contents of the auger were emptied into a cotton soil-sampling bag and transported to the University of Pretoria.Soil samples which were used for the re-examination were stored in a dry, dark place at ambient temperatures until needed.
Samples were collected in January, April, July and September 2001, representing the summer, autumn, win ter and spring soil seed bank, respectively.All soil sam ples were examined by the seedling emergence method immediately after collection.The results of this exami nation will be referred to as the seasonally germinable seed bank.The remaining soil of 15 randomly selected samples per habitat type was used in flotation studies (Kellerman 2004), and the remaining soil of the other 45 soil samples collected on one date per habitat type were subjected to a re-examination in September 2001.The rationale for the re-examination was, that if newly shed seeds had an after-ripening requirement, they would not germinate immediately after collection and would there fore not be detected by the seedling emergence method.By allowing an after-ripening period of a few months and re-examining duplicate samples in September, when temperatures were thought to be optimal for seed ger mination, the problem of an after-ripening requirement could be overcome.The results of this re-examination will be referred to as the persistent fraction of the soil seed bank.In arid regions where dormancy mechanisms are common for many species, such a re-examination indicates the size of the potentially-germinable seed bank (De Villiers et al. 2004).
The seedling emergence method was used to deter mine the seed bank size and species composition of each soil sample.Plastic pots (100 x 100 x 120 mm) were filled with finely ground quartz and topped with 100 cm^ of soil from a sample.The pots (n = 60 per site per examination time) were placed in an uncontrolled greenhouse and watered daily with tap water and fort nightly with Amon and Hoagland's complete nutrient solution (Hewitt 1952).Once a week, for a period of three months, all newly emerged seedlings were marked using toothpicks.Duplicate samples to determine the persistent seed bank fraction were investigated in the same manner in September 2001.Seedlings were iden tified as soon as possible and, once identified, removed from the pots to prevent contamination by self-seeding.Unidentified seedlings were left to mature for later iden tification.Voucher specimens are housed in the H.G.W.J. Schweickerdt Herbarium and nomenclature follows Germishuizen & Meyer (2003).
The percentage correspondence in species composi tion between the different habitat types and between the seasonally germinable and persistent fractions of the seed bank was calculated by means of Jaccard's Similarity Index (IS^) which reads as follows: where c is the number of species common to both habitats or examination dates, a is the number of species restricted to one habitat or examination date, and b is the number of species restricted to the other habitat or exam ination date (Mueller-Dombois & Ellenberg 1974).
Data were analysed by using a one-way analysis of variance and Tukey's post-hoc test in the Statistica 7 com puter program (StatSoft Inc., Tulsa, Oklohoma, USA).

Seed bank size
In general, the Licuati forest (Figure lA) and Licuati thicket (Figure IB) soils contained the smallest num ber o f germinable seeds.Mean seed densities for the Licuati forest ranged from a low of 1 067 seeds m'^ in April to a high of 1 950 seeds m'^ in January and the Licuati thicket from 1 050 seeds m'^ in April to 3 200 seeds m'^ in January.In both cases the seasonal variation in seed bank size was statistically significant (p = 0.013 for Licuati forest; p < 0.001 for Licuati thicket).The Licuati forest and thicket seed banks showed a similar trend in seasonal variation in that they had the smallest seed numbers present in the soil during autumn, gradu ally increasing towards spring and reaching the highest seed densities in spring or summer.The Licuati thicket produced a significantly larger (p < 0.001) seed bank in summer than the Licuati forest, but seed bank size was similar in other seasons (p > 0.05 in all cases).When the Licuati forest soil samples were re-examined in spring, seed bank size was within 10% of that of the seasonally germinable seed bank.In the case of the Licuati thicket it was found that the autumn and winter seasonally ger minable seed banks produced significantly larger seed numbers than the re-examination in spring (p = 0.006 for autumn comparison; 0.0008 for winter comparison), whereas the difference was not significant for the sum mer seed bank (p < 0.001).
In the forest/grassland ecotone, changes in the size of the seasonally germinable seed bank were damped (p = 0.597 over four seasons) (Figure IC) and seed density ranged from 4 034 seeds m'^ in July to 4 667 seeds m' ^ in September.When summer and winter collected soil samples were re-examined in spring, higher seed densi ties were encountered; however, these differences were not significant.The re-examined autumn soil samples yielded significantly less seeds (p = 0.004).
The seasonally germinable seed density of the grass land vegetation remained fairly constant from summer through autumn to winter, but showed a marked decline in spring (p = 0.002 over four seasons) (Figure ID).The number o f germinable seeds in the grassland soil almost halved from a high o f 5 567 seeds m'^ in January to a low o f 2 734 seeds m'^ in September.The seasonally ger minable summer and autumn seed banks from the grass land soil produced significantly more seedlings than the duplicate analysis in spring (p < 0.001 for summer comparison; p < 0.001 for autumn comparison).This persistent fraction o f the seed bank of the grassland soil remained relatively constant (p = 0.897) throughout the year.
Seasonal variation in seed densities in the woodland soils varied significantly (p < 0.001) and ranged from 2 167 seeds m'^ in September to 6 467 seeds m'^ in July.In contrast to the other vegetation types, the largest sea sonally germinable seed bank in the woodland soils was present during winter (Figure IE).The re-examination in spring produced significantly less seedlings than the sea sonally germinable seed banks o f all seasons (p < 0.001 for summer comparison; p <0.001 for autumn compari son; p = 0.039 for winter comparison).

Species composition
In total, 52 genera and 83 taxa were identified in the soil samples, together with a few unidentified specimens (Appendices 1-5).Many seedlings died while still in the cotyledon stage before they could be positively identi fied.These unidentified seedlings are indicated in the tables as mortalities.In general, the Licuati forest and thicket seed banks had the lowest species richness per sampling time.
The largest number o f taxa consistently emerged from the woodland seed bank.The highest number o f taxa at a single sampling time was 35, recorded from the spring woodland soil seed bank.The re-examination of soil samples of the summer, autumn and winter collections in spring generally produced species richness counts either the same or less than the examination immediately after collection.However, the decrease in richness could be the result of the re-examination being based on 45 soil samples as opposed to 60 for the initial examination immediately after collection.
The seed bank flora o f the Licuati forest and thicket differed vastly from those o f the other three habitat types.Thirty-three taxa comprising nine grass, four sedge, 18 herbaceous and two woody species were identified in the Licuati forest seed bank (Appendix 1).The Licuati thicket soils produced 25 identified taxa including nine grass, four sedge and 12 other herba ceous species (Appendix 2).Diagnostic species o f the Licuati forest and thicket seed banks included Crassula c f expansa, Cyperus dubius, Eragrostis cf.moggii, Leptochloa c f uniflora and Panicum laticonum.Only two woody species were encountered in the seed banks o f the Licuati vegetation types.The seed bank flora was often dominated by a single species and seed den sities in the Licuati forest seed bank ranged from 17 to 1 222 seeds m'^ for individual species at a particular sampling time and for the Licuati thicket from 17 to 2 822 seeds m Floristic composition of the ecotone soil seed bank consisted of 45 taxa including 15 grass, six sedge, 23 herbaceous and one woody species (Appendix 3).The seed bank flora of the ecotone was dominated by the grass species Perot is patens, constituting 58.4% of the total number of seedlings.Other prominent species included Bulbostylis burchellii, B. hispidula, Cyperus chersinus, Kohautia virgata, Phyllanthus c f parvulus and Setaria sphacelata var.sphacelata.
In the woodland soil seed bank the emergence of 61 species including 20 grass, seven sedge, 33 herbaceous and one woody species was recorded (Appendix 5).Prominent grass species obtained from the soil samples were Aristida stipitata var.stipitata, Brachiaria chusque- Species such as Conyza albida, Gamochaeta pennsylvanica, Hypochaeris radicata, Oxalis cf.semiloba, Pseudognaphalium luteo-album and Sonchus asper occurred in relatively large densities in the seed banks of all the investigated habitat types.Jaccard's Similarity Index indicated a 62% correspon dence in species composition between the Licuati forest and thicket seed banks (Appendix 6).The species com position of the forest/grassland ecotone and grassland seed banks also showed a high degree of similarity (58%, Appendix 6).There was a low correspondence of 29% between the Licuati forest and the forest/grassland eco tone in spite of their close spatial association.The wood land seed bank was more closely related to the grassland seed bank in composition than to the Licuati vegetation types.In general there was a low correspondence in spe cies composition between the seasonally germinable seed bank and the re-examination in spring (Appendix 7).

Seed bank size
Several studies have indicated that forest seed banks are relatively small to almost nonexistent in both size and species richness (Warr et a l 1993;Crawford & Young 1998).Mature, dry, tropical forests generally produce very low seed bank densities and should there be higher seed numbers present in soils underlying dry forest vegetation, it could be ascribed to the abundant seed rain o f pioneer species characteristic o f early successional stages (Matlack & Good 1990;Alvarez-Buylla & Garcia-Barrios 1991;Rico-Gray & Garcia-Franco 1992;Falinska 1998;Jankowska-Blaszczuk et al. 1998;Arevalo 8c Femandez-Palacios 2000;Guariguata 2000).Data from Licuati forest and thicket soils collected in the present study agreed well with these findings in that they produced the lowest seed densities o f all the examined habitat types.Examples of mean seed densities obtained in other forest seed bank studies by the seedling emer gence method include 265 to 2 910 seeds m'^ (Matlack 6 Good 1990), 203 to 5 613 seeds m ^ (Brown 1992), 330 to 3 437 seeds m'^ (Jankowska-Blaszczuk & Grubb 1997), 156 to 4 148 seeds m'^ (Falinska 1998), 610 to 7 009 seeds m'^ (Halpem et al. 1999) and 137 to 6 920 seeds m'^ (Olano et al. 2002).Seed densities reported for the Licuati vegetation types (1 050 to 3 200 seeds m'^) are in the same order o f magnitude as those mentioned above for the other forest types.However, Kellerman (2004) warned that these values, obtained with the seed ling emergence method, could have greatly underesti mated the true size o f the seed bank.The most notewor thy difference in the seed bank size between the Licuati forest and thicket vegetation types was that the Licuati thicket produced a significantly larger soil seed bank than the Licuati forest in summer.This could reflect dif ferences in the floristic composition, reproductive strate gies, timing o f seed fall, seed germination requirements and seed dispersal efficiency between these two vegeta tion types.
Compared with the forest and woodland soils, the forest/grassland ecotone soil produced a large seed bank almost year round, implying that seasonal variation was of lesser significance.The same trend in temporal variation was observed in grasslands studied by Coffin & Lauenroth (1989), Milberg & Hansson (1993) and Kalamees & Zobel (1997).The grassland seed bank sampled in the Tembe Elephant Park only partly followed this trend being rela tively constant for the largest part o f the year, but show ing a marked decline in spring.The persistent fraction of the seed bank o f the grassland soil remained remarkably constant throughout the year.Mean seed density varied from 4 034 to 4 667 seeds m'^ in the forest/grassland ecotone and from 2 734 to 5 567 seeds m'^ in the grass land, which compares well with reported grassland seed densities o f 122 to 2 748 seeds m'^ (Coffin & Lauenroth 1989), 2 580 to 10 060 seeds m'^ (Milberg & Hansson 1993) and 1 421 to 2 567 seeds m'^ (Kalamees & Zobel 1997).Seed bank studies o f South Afncan grasslands by O ' Connor & Pickett (1992) and Adams (1996) found mean seed densities varying between 300 and 10 000 seeds m'^.
The woodland seed bank differed from the for est/grassland ecotone and grassland seed banks in that it produced distinctly more germinable seeds in winter with greatly reduced seed densities observed in spring.Mean seed density varied from 1 734 to 6 467 seeds m'^.Dougall & Dodd (1997) and Garcia-Nunez et al. (2001) reported similar seed densities in their studies of neo tropical savanna vegetation with mean seed densities of 897 to 9 100 seeds m'^.The relatively smaller seed densi ties recorded from both the grassland and woodland soils in spring might have been the consequence of germina tion induced by early rainfall in September before the soil collection.Rainfall data for the Sihangwane Weather Station in Tembe Elephant Park for September 2001 was 22 mm, which would have been sufficient to trigger the germination o f many seeds.

Species composition
Each habitat's seed bank was characterized by its own species composition although the Licuati forest and thicket seed banks showed a large degree of similarity, as did the forest/grassland ecotone and the grassland seed banks (62% and 58% respectively).The relatively low (29%) similarity between the Licuati forest and the forest/grassland ecotone stresses the abrupt transition in species composition from the forest to the surrounding edge.Species composition showed large seasonal varia tion within a specific habitat type.Thompson & Grime (1979), Roberts (1986) and Morgan (1998) drew similar conclusions.The largest similarity in species composi tion was found between the summer and autumn seed banks.
Dry, tropical forest types, such as the Licuati forest and thicket, are generally characterized by high species richness in the standing vegetation (Murphy & Lugo 1986;Swaine 1992;Matthews et al. 2001, Gaugris et al. 2004).However, tropical forest seed banks are usually very small or almost nonexistent and a large portion of the viable seeds that do occur in forest soils, belong to pioneer species (Roberts 1981;Thompson 1985;Rico-Gray & Garcia-Franco 1992;Skoglund 1992;Bigwood & Inouye 1998;Jankowska-Blaszczuk et al. 1998).The similarity in species composition between the seed bank and the aboveground flora in the Licuati forest was reported to be only 7.8% (Kellerman 2004).An analysis of the floristic composition of both the Licuati forest and thicket soils showed that the soil seed bank was com posed primarily o f grasses, sedges, annuals and short lived perennial species, supporting the evidence from other studies that dry, tropical forest climax species do not produce persistent seed banks.
It is generally believed that the Sand Forest is unable to regenerate after a major disturbance and this has led to many concerns about the long-term survival of this vegetation type.The lack of seeds of forest or thicket canopy species in the soil seed pool o f these vegeta tion types could be one of the contributing factors why the Licuati vegetation does not regenerate readily after major disturbances.In the present study the seeds of the prominent canopy species were absent at all examination times, indicating that the seed bank of these species is either very short-lived or that the germination require ments of the seeds were not met by the seedling emer gence method applied.If the seed banks of canopy spe cies are transient, as most studies on forest seed banks seem to indicate, regeneration would have to depend on seed dispersal from undisturbed sites.
The re-examination of the forest and thicket soils revealed a few important trends.The summer seasonally germinable seed bank was dominated by weedy annual species such as species of Conyza, Sonchus cf.asper, Gamochaeta pennsylvanica and Pseudognaphalium luteo-album.These species apparently do not per sist well in the seed bank, because their numbers were greatly reduced in the re-examination of duplicate sam ples in spring.In contrast, the re-examined samples were dominated by other species e.g.Leptochloa c f uniflora {Licuati forest) and Panicum laticonum (Licuati thicket).These grass species dominating in the spring re-exami nation apparently require some after-ripening to break seed dormancy.
The seasonally germinable seed bank of the for est/grassland ecotone was dominated by the grass spe cies Perotis patens and sorrel Oxalis c f semiloba.In the grassland seed bank, the relative contributions of the species differed among seasons.The most important species in the seasonally germinable seed bank were Bulbostylis hispidula, Conyza albida, Oxalis cf.semiloba and Sonchus c f asper.The re-examination in spring of both the ecotone and grassland soils showed an increase in the seed density of Perotis patens, indicating that the seeds of this species require some after-ripening for optimal germination.A study to investigate the similar ity between the seed bank and aboveground flora found a 44.4% similarity for the grassland (Kellerman 2004).This value is substantially higher than that of the Licuati thicket.
The woodland vegetation in the Tembe Elephant Park is composed of an upper tree layer and a prominent her baceous or grass layer.The woodland soils examined in this study produced the highest number of species.Interesting though, was that the woodland soils produced the highest richness from the smallest seed density as was observed from the spring seed bank data.Despite the large number of species, only one woody species was recorded from the woodland soil seed bank.
In conclusion, with the exception of the forest/grass land ecotone, the seed banks of all habitat types inves tigated in this study, showed pronounced seasonal vari ation.The Licuati forest and Licuati thicket seed banks had the lowest seed densities and also the lowest species richness.In general, grass and sedge species comprised more than 40% of the seed bank flora that emerged from the soil samples.The remaining species were mostly annual and perennial forbs, with hardly any evidence of woody species.

Dir. Re
Apr.

Dir. Re
Jul.

Dir. Re
Sep. Species Dir. Jan.

Dir. Re
Apr.

Dir. Re
Jul.

A ncylobotrys peter
FIGURE 1.-Seed bank size (mean number o f seeds m'^) as obtained from soils collected in five selected habitat types in Tembe Elephant Park, when examined directly after collection, and re-examined in September.A, Licuati forest; B; Licuati thicket; C, forest/grassland ecotone; D, grassland; and E, woodland.