Introduction
Seed mass, a key ecological trait that affects many aspects of plant ecology (Moles et al., 2005a, b; Mason et al., 2008), has great influences on the regeneration strategies of plants, including seed output for a given amount of energy, seed dispersal and seedling survival (Leishman et al., 2000). Variation in seed mass reflects the fundamental trade-off between seed number and seed mass (Henery & Westoby, 2001) and between seed mass and persistence in the seed bank (Thompson et al., 1993). An increasing body of evidence has shown that large-seeded species produce less seeds than those bearing small seeds (Henery & Westoby, 2001; Moles et al., 2004). Compared to small-seeded species, large-seeded species are more likely to produce large seedlings that are supposed to survive better than small seedlings under a variety of hazardous environments (Armstrong & Westoby, 1993; Leishman & Westoby, 1994a, b; Burke & Grime 1996; Westoby et al., 1996, 2002; Harms & Dalling, 1997; Leishman et al., 2000; Dalling & Hubbell, 2002; Moles & Westoby, 2004; Dainese & Sitzia, 2013). Despite the advantages associated with large seeds, seed masses of present-day species have been observed to range over 11.5 orders of magnitude, from the 0.0001-mg dust-like seeds of orchids to the 20-kg seeds of the double coconut (Michelle et al., 1995). Therefore, to reveal the internal mechanism and influencing factors of the changes in seed mass will help us to better understand the ecological history of plants.
As proposed by the leaf-height-seed (LHS) scheme of plant ecology strategy (Westoby, 1998), plant height and leaf area are closely correlated with seed mass. As a crucial component of a plant species’ ecological strategy, plant height not only determines a plant’s ability to compete for light but also a species’ carbon gain strategy, which is supposed to play an important role in another life-history trait, seed mass (Moles & Leishman, 2008). A pioneering study by Levin (1974) found that the mean seed mass of 832 plant species increase along the growth form: herbs, shrubs, vines, shrubby trees, and trees. Leishman et al. (1995) showed that seed masses are consistently correlated with plant height across 1659 species, representing a worldwide flora. Carly et al. (2009) found positive linear correlations between plant height and seed mass of 15 species at the community level in the northeast Galilee region of Israel. Moles et al. (2004) analyzed the trait data of 2026 species across 150 families and revealed a positive correlation between seed mass and plant height, representing large-scale evidence for the relationship between seed mass and plant height. However, Grime et al. (1997) found no significant correlation between plant height and seed mass across 43 common British species. Thompson & Rabinowitz (1989) analyzed 816 plant species around Sheffield and found significant relationships between seed mass and plant height within some families (Poaceae, Caryophyllaceae, Asteraceae and Fabaceae), but not in other taxa like Scrophulariaceae, Apiaceae, Lamiaceae, Calophyllum ,Pinus , and Quercus . In a southeastern Sweden flora, seed mass was only marginally correlated with plant height of 126 species (Bolmgren & Cowan, 2007). Rees (1996) analyzed 382 species of Sheffield flora and found that the relationship between seed mass and plant height is inconsistent and dependent on dispersal modes. Therefore, much uncertainty still remains to be tackled possibly due to the limitation of plant species sampled in previous studies, although plant height has been considered one of the strongest correlates of seed mass (Leishman et al., 1995; Moles et al., 2004).
As the main organ of plants that contributes to photosynthesis (Bazzaz et al., 2000), leaves act as a key determinant of the amount of energy available for reproduction (Wright et al., 2004). Although leaves may vary in their traits (e.g., area, morphology, anatomy, physiology and N nutrition) in response to growing conditions (Givnish, 1987; Witkowski & Lamont, 1991; Ackerly & Reich, 1999; Cornelissen et al., 2003; McDonald et al., 2003; Xu et al., 2009; Milla & Reich, 2011), a strong connection between total leaf mass and net annual reproductive biomass has been observed (Niklas & Enquist, 2002, 2003). Therefore, the ecological significance of leaf traits may relate to resource capture in productive organs, implying that leaf area and seed mass should be positively correlated (Westoby & Wright, 2003). Midgley & Bond (1989) found that leaf area was positively correlated to cone size in 18 species from the Leucadendron genus in South Africa. Hodgson et al. (2017) also found positive linear relationships between leaf area and seed mass of 2400+ species from England and Spain. Although it was not specifically stated in the studies by Laughlin et al. (2010), seed mass appears to be positively correlated with leaf area of 133 plant species in northern Arizona, USA. Rather than a linear relationship, Cornelissen (1999) showed a triangular relationship between leaf area and seed mass of 58 woody species from Europe. Recently, Santini et al. (2017) showed that the triangular relationship also holds for 401 annual plants belonging to 37 families from the United Kingdom. However, Westoby & Wright (2003) failed to find the triangular relationship between leaf area and seed mass as reported by Cornelissen (1999), indicating that the pattern seems not universal between seed mass and leaf area.
In addition, seed mass is not independent of growth form, which is often a predictor of other plant traits (Moles et al., 2005a, b). Plant growth form, like seed mass, may also be phylogenetically controlled (Li et al., 2017). Evidence has shown that woody plants are more likely to have larger seeds, while non-woody species are more likely to produce small seeds (Jurado et al., 1991). Therefore, the phylogenetic conservatism of plant growth form might have an indirect impact on the variations in seed mass. Furthermore, genome size appears to be one of the most studied factors that are related to variations in seed mass. Although there is no significant linear regression relationship between genome size and seed mass across 1222 species from 139 families and 48 orders of seed plants, Beaulieu et al. (2007) found that species with very large genome sizes never had small seeds. Therefore, apart from the influence of plant height and leaf area, phylogeny, growth form and genome size may also contribute to seed mass variations.
Phylogenetic conservatism in plant traits has been well studied (Wiens et al., 2010; Baskin & Baskin, 2014; Cornwell et al., 2014; Tozer et al., 2015) and such studies are helping to illuminate the role of the evolutionary past in determining the characteristics of species. Seed mass has been accepted as an ecologically important trait phylogenetically constrained within local floras. This may also be true for plant height and leaf area. Therefore, it should be obligatory to extract variations in plant traits associated with phylogeny, before analyzing relationships between seed mass and other plant ecological attributes, e.g., growth form, plant height, and leaf area. However, the potential influence of phylogeny on the leaf-height-seed (LHS) plant ecology strategy scheme has not previously been well evaluated (Cornelissen, 1999; Laughlin et al., 2010; Hodgson et al., 2017).
Despite previous data on the relationship between plant traits across the world, published literature failed to incorporate phylogeny and plant traits into the analysis of the variations in seed mass (Chase & Pippen, 1990; Mustart & Cowling, 1992; Lord et al., 1995; Kang & Primack, 1999; Zhang et al., 2004; Vandelook et al., 2018). The rapid accumulation of databases on plant traits provides us an ideal opportunity to illustrate a general pattern of the relationship between plant traits, which helps us to have a better understanding of the leaf-height-seed (LHS) plant ecology strategy scheme. In the present study, we first used phylogenetic partial R2s (Ives, 2019) to tease apart the effects of multiple plant traits (plant height, leaf area, genome size, growth form and leaf N) and phylogeny, to quantify extent to which they contribute to variations in seed mass of plant species when each predictor variable and the phylogeny is removed one-by-one.