1. Previous studies have shown that among British species, seeds that persist in the soil tend to be small and compact compared with non-persistent seeds. We tested whether or not this pattern is repeated among 101 Australian species, from a range of habitats.
2. Seed mass was plotted against variance of seed dimensions, across all species. Species with persistent seeds were found across the whole range of seed mass (0·217–648·9 mg) and variance (0·0000–0·2497), providing no evidence for a critical mass or variance which separated persistent from transient seeds.
3. We tested whether or not divergence within individual clades between persistent and transient seeds was associated with increased seed mass or seed dimension variance, using phylogenetically independent contrasts (PICs). There was no consistent tendency found.
4. Thus for Australian species, persistent seeds were not smaller or more compact than transient seeds when compared across all species or when compared using PICs. Presumably the natural history of burial and disturbance operates differently in British and Australian habitats.
Thompson, Band & Hodgson (1993) showed that among 97 species of the British flora, seeds which are persistent in the soil tended to be small and compactly shaped, while seeds that are transient in the soil were larger, flattened or elongate. Importantly, they suggested that the generality of the suspected underlying mechanism (i.e. ease of burial) implied that the method might be applicable to floras outside north-west Europe.
Here we assess whether or not the pattern reported by Thompson et al. (1993), that seeds that are persistent in the soil tend to be small and compact, is repeated among 101 Australian species from a range of habitats.
Materials and methods
Data on seed mass, seed dimensions and persistence in soil were collected for 101 species of the Australian flora, from 60 genera and 29 families, which occur in a range of habitats. Habitats included heaths and woodland of the Sydney region and Western Australia, semiarid woodlands of New South Wales, Queensland and the Northern Territory, tall forest of Victoria and woodland of the subalpine region of New South Wales and Victoria. The 101 species included monocots, herbaceous dicots, shrubs and trees. The species list also included four cosmopolitan species, originally native to Europe. Species were selected on the basis that information on dormancy was available and seed material was accessible.
Information on seed dormancy was recorded from the literature (e.g. Cunningham et al. 1981; Pate & McComb 1981; Langkamp 1987; Irons & Quinlan 1988; Bell, Plummer & Taylor 1993). Each species was classified into one of four classes: immediate germination, seasonal dormancy only (between seasons or <5 years), dormant (>5 years) or fire-promoted germination. This information was then used to classify species into those with seeds persistent in the soil and those with seeds transient in the soil. The first two classes of seeds were considered as ‘transient’ in the soil seed bank, while the latter two were considered as ‘persistent’.
Seed material for each species was obtained from field collections or commercial seed companies. For each species, a minimum of 10 seeds was weighed (to μg precision on a Cahn microbalance; Cahn Instruments Inc., CA, USA), and seed length, width and depth were measured (to μm precision using Vernier calipers). Seed variance was then calculated following the methods of Thompson et al. (1993), as the variance of seed length, width and depth after transforming all values so that length was unity, to give a measure of seed shape. Seed variance values close to zero represent near-spherical shapes. Seeds were measured as the seed coat, embryo and endosperm only, with dispersal structures removed.
Seed mass ranged from 0·217 mg to 648·9 mg, while variance ranged from 0·0000 to 0·2497 (data in Appendix 1). This is a wider range of seed mass than in the British flora dataset used by Thompson et al. (1993), which was confined to herbaceous species.
The pattern reported by Thompson et al. (1993) can be expressed as the boundary indicated by the dashed line in Fig. 1. In the British herbaceous flora dataset, all species with seed mass low enough and variance of dimensions low enough to lie within the boundary, had persistent soil seed banks. The data points in Fig. 1 are for Australian species, with closed symbols for persistent seed banks. Evidently the British rule does not apply in Australia. Species with dormancy are not confined within the boundary, but are found across the whole range of seed masses (from 0·50 mg to 648·9 mg) and the whole range of seed variances (0·0000–0·2497). This is true both of species with bet-hedging dormancy and of species with dormancy that defers germination until after fire. In addition, seeds that are transient in the soil (immediate germination or seasonal dormancy) are also found in all areas of the graph. Thus, not only does the particular seed mass and shape cut-off between persistent and transient species of the British flora not hold true for species of the Australian flora but also no such line defining a seed size and shape relationship could possibly be drawn to separate seeds of persistent from transient species within the Australian flora.
It is not the case that species with a mechanism to ensure burial of their seeds (elaiosomes or hygroscopic awns, see Appendix 1), are the only persistent species found outside such a seed size/shape position in Fig. 1. Only seven of the 75 persistent species have such a burial mechanism, and they are scattered across all regions of the graph.
The 101 species came from a diversity of habitats and climates. Only two habitats had sufficient numbers of species such that both transient and persistent seeds were adequately represented to allow within-habitat analysis of the seed size/shape pattern in relation to seed persistence. There was no evidence for any seed size/shape relationship separating transient from persistent seeds in these two habitats: the semiarid woodlands or heaths and woodlands of Western Australia.
The relationship between dormancy and seed size and shape found by Thompson et al. (1993) for the British flora was consistent both within and between families. Therefore we considered the possibility that even though the overall relationship was absent in the Australian flora, nevertheless divergence within individual clades between persistent and non-persistent seed banks might tend to be associated with either increasing seed mass or increasing seed dimension variance. To investigate this possibility, we isolated from the Australian dataset phylogenetic divergences that were contrasted for seed persistence. Each side of a divergence could be an individual species or an average across several. Each contrast was phylogenetically independent of the other contrasts. Each such phylogenetically independent contrast (PIC) represents a separate episode of past evolutionary divergence for seed persistence. We adopted the flowering plant phylogeny described by Bremer, Bremer & Thulin (1996), which is designed to represent a consensus of present knowledge of the angiosperm phylogenetic tree based on molecular sequence information as well as other sources. The results are shown in Fig. 2, where each PIC is denoted by a letter (A to M), with the uppercase letter showing the mean seed mass and variance for persistent species and the lowercase letter the mean seed mass and variance for transient species. Each line leading from an uppercase to a lowercase letter represents an evolutionary divergence from persistent to non-persistent seed banks. Clearly in the Australian flora there is no consistent tendency for those lines to lead towards higher seed mass or higher seed dimension variance.
The pattern found by Thompson et al. (1993) whereby seed persistence in the soil could be predicted from seed size and shape, was not repeated across 101 species of the Australian flora, encompassing a wide range of growth forms and habitats. In Australia, persistent seeds were not smaller and more compact than transient seeds, when compared across all species or when compared using phylogenetically independent contrasts.
Thompson et al. (1993) found that the cut-off between transient and persistent seeds differed slightly between fruits and seeds, with persistent seeds being more compact than persistent fruits. In this study, we examined only seed masses and shapes. However, not only did we find that the seed size and shape cut-off for British seeds was not true for Australian seeds, we found no evidence whatsoever for any seed size, shape and persistence relationship among Australian species.
This study has two main results. First, the method for predicting seed persistence in the soil based on British species and reported by Thompson et al. (1993), cannot be extended to the Australian flora. Presumably conditions of burial and disturbance operating in Australian environments are different from those operating in Britain. The British flora is largely herbaceous and consists of many species which have been able to colonize readily following deglaciation throughout the Quaternary. Possibly, species with dormant seeds which are small and compact have been particularly successful in such a context. In contrast, in the Australian habitats used in this study, fire is the major form of disturbance. Many of the seeds which are persistent in the soil of Australian habitats are hard-seeded, such as members of the Fabaceae, Mimosaceae and Caesalpinaceae. These hard-seeded persistent species typically have a greater seed mass than the non-hard-seeded persistent species (i.e. 3–30 mg compared with < 3 mg). However, it is not solely these hard-seeded persistent species which are responsible for the differences in the seed size/shape pattern between the British and Australian species. Among the Australian non-hard-seeded persistent species, seed variance varies across the entire range. Thus these non-hard-seeded persistent species may be small in mass like their British counterparts, but they are not typically compact in shape.
Second, this study has highlighted the usefulness of extending results from comparative plant ecology to other floras, having different climates and vegetation histories. Although there is a theoretical basis for expecting a trade-off between seed size and dormancy (Venable & Brown 1988; Philippi & Seger 1989), such a trade-off would not necessarily be the main or the only effect shaping the relationship in all environments. By extending comparative analyses to different floras, habitats and climates, we can gain a stronger sense of the relative strengths of different trade-offs and relationships among plant attributes. Obvious questions arising from the results of this study are: what floras share the natural history characteristics of British habitats such that small, compact seeds tend to be associated with persistent seed banks? What floras share the natural history characteristics of Australian habitats such that there is no relationship between seed size, shape and persistence? And do areas with other natural history characteristics result in different seed size/shape and persistence patterns? We hope that researchers familiar with further different floras will take up the challenge.
M. R. Leishman & M. Westoby
We thank Sonia Rousseau for technical assistance. This work was supported by the Australian Research Council and is contribution number 262 from the Centre for Biodiversity and Bioresources.
Data on species name, family, seed persistence, seed mass (mg) and seed variance for the 101 Australian species measured. Species names follow Hnatiuk (1990). Seed persistence was determined from literature sources such as Langkamp (1987); Pate & McComb (1981) and Bell et al. (1993). Persist 1, immediate germination; 2, seasonal dormancy only; 3, dormant; 4, fire-promoted germination. Species marked with an asterisk have a burial mechanism (elaiosome or hygroscopic awn). Seed mass was measured to μg precision on a Cahn microbalance, using a minimum of 10 seeds for each species. Variance was calculated as the variance of seed length, width and depth, after transforming all values so that length was unity. The letter in the PIC column denotes which phylogenetically independent contrast the species data were used for (see Fig. 2).