Traits of recalcitrant seeds in a semi-deciduous tropical forest in Panamá: some ecological implications

study site and species

This study was conducted on the Barro Colorado Nature Monument (BCNM), Republic of Panamá (9°10′ N, 79°51′ W). Vegetation on the BCNM consists of semi-deciduous tropical forest, and has been described in detail elsewhere (Leigh, Rand & Windsor 1982): nomenclature follows the Flora of Panama Checklist (D’Arcy 1987). Rainfall on the BCNM averages 2600 mm year⁻¹, with a pronounced dry season between January and April (Dietrich, Windsor & Dunne 1982).

Ripe fruits/seeds, at the point of natural dispersal, were collected from 226 tree, liana and shrub species (see Appendix 1) between 1985 and 1989. Fleshy fruits were cleaned by removing the fleshy pulp within 2 days of collection: no cleaning was necessary for wind-dispersed seeds and those in dry, dehiscent pods. For each species the month of seed collection was recorded and seed dry mass was determined by drying c. 10 cleaned seeds per species (with fruit tissue removed) at 60 °C for 3 days.

Data on seed desiccation tolerance was collated from Release 6 of the Royal Botanic Gardens Kew's online Seed Information Database (SID; Flynn, Turner & Dickie 2004). Seed responses to desiccation are divided into three categories in SID: orthodox (desiccation-tolerant); recalcitrant (desiccation-sensitive); and intermediate. Intermediate seeds have seed storage characteristics that are intermediate between those of orthodox and recalcitrant taxa and account for only ≈2% of SID. Species in the data set were assigned to these seed storage categories, resulting in 189 orthodox, 36 recalcitrant and 0 intermediate taxa.

germination experiments

Where seed numbers allowed, germination tests were conducted on each seed collection within the growing house in the laboratory clearing on Barro Colorado Island (BCI; part of the BCNM). Following collection, seeds were stored at 25 °C and sown within 2 days. For germination experiments, seeds were sown in plastic or fibre (peat) pots containing unsterilized soil collected from within the forest. The number of seeds sown per pot was determined primarily by seed size, and ranged from one per pot in the case of the larger-seeded species (e.g. Prioria copaifera) to 200 for the smallest seeded species (e.g. Conostegia spp.). For each species between 20 and 200 seeds were planted. Seeds were sown on the soil surface in an attempt to mimic natural dispersal by wind, being dropped by animals or deposited in faeces. However, a few of the larger-seeded species are actively buried by scatter-hoarding rodents; we did not take this into account. As a number of the species in this study can be classified as pioneers (sensuSwaine & Whitmore 1988) and require high-light conditions for germination (Daws et al. 2002a), pots were split between the upper and lower shelves in the growing house to mimic conditions within canopy gaps (upper shelf) and closed understorey sites (lower shelf) (Garwood 1983; Molofsky & Augspurger 1992). On the upper shelf, irradiance levels averaged 18·5 ± 0·7% (±1 SE) full sunlight vs 2·3 ± 0·2% on the lower shelf. These values are typical of medium-sized gaps and intact understorey sites, respectively, on BCI (Daws et al. 2002a; Pearson et al. 2002).

Pots were watered daily or as required (Garwood 1983), and germination was scored at weekly intervals until either (1) all seeds had germinated; (2) all remaining seeds were badly damaged or had disappeared from the pots (restricted to larger-seeded species); or (3) >3 months had passed with no further germination since the last germination event. Consequently, many germination tests extended up to 3 years. Germination was scored as visible seedling emergence.

For each taxa, the mean time to germination (MTG) on both upper and lower shelves of the growing house was calculated using the following equation:

where n is the number of seeds germinated between scoring intervals; d the incubation period in days at that time point; and N the total number of seeds germinated in the treatment (Tompsett & Pritchard 1998).

Determination of resource allocation to defence

For each species, a minimum of eight individual seeds (dispersal unit) were dissected into their component parts: endocarp/testa and embryo/endosperm. These component parts were subsequently dried at 103 °C for 17 h (ISTA 2004) followed by mass determinations. To calculate the allocation to defence (seed-coat ratio, SCR), the ratio of the mass of covering structures (endocarp and testa) to the mass of the total dispersal unit was determined (Grubb & Burslem 1998; Pritchard et al. 2004).

statistical analysis

One-way anova implemented in minitab 13 (Minitab Inc., State College, PA, USA) was used to test for a relationship between desiccation sensitivity and seed mass, MTG or SCR. In all analyses, seed mass was logarithmically and SCR arc-sine transformed to ensure homoscedasticity. For both cross-species and phylogenetic analyses involving MTG, two separate analyses were performed using MTG in the shade (MTGS; lower shelf) and in the light (MTGL; upper shelf). MTG was transformed using a Box–Cox transformation with the optimum value of λ determined using minitab 13. This resulted in λ values of −0·224 and −0·225 for MTGS and MTGL, respectively. The effect of either MTG or SCR on desiccation sensitivity while removing the effect of seed mass was tested using ancova in minitab 13. In addition, sign tests were used to test the null hypothesis, for desiccation-sensitive and desiccation-tolerant species separately, of no difference between MTG in light and in shade.

For each species the timing of dispersal was classified as dry or wet season. This classification was based on whether dispersal occurred in January to April, inclusive (dry season) or May to December (wet season) (Daws et al. 2005). Subsequently, for desiccation-tolerant and desiccation-sensitive species, a χ² test of association was used to test the null hypothesis of no association between response to drying and timing of dispersal.

Phylogenetically independent contrasts (Felsenstein 1985; Pagel 1992) were used to analyse the relationship between desiccation sensitivity and seed mass, MTG and SCR. This approach is based on the logic of comparing pairs of species within a phylogeny that share an immediate common ancestor. The null hypothesis is that there is no correlation between changes in traits at the nodes. The package caic (Purvis & Rambaut 1995) was used to generate contrasts. Within the phylogeny we assumed that all branch lengths were the same: analyses of simulated data sets suggest that equal branch lengths may perform better than estimated branch lengths (Purvis, Gittleman & Luh 1994). For these analyses, the Brunch procedure, designed for discrete predictor variables, was used (Purvis & Rambaut 1995).

For analyses testing whether the transition from desiccation-tolerant to desiccation-sensitive seeds is associated with changes in either MTG or SCR while holding the effect of seed mass constant, contrasts between mass and MTG or SCR were calculated using the Crunch procedure in caic, designed for continuous data (Purvis & Rambaut 1995). Subsequently a linear regression, forced through the origin, was fitted to the contrasts. A linear regression with the same slope was then fitted to the raw data for mass and MTG and for mass and SCR, and the contrasts from the line recorded. Changes in these contrasts associated with seed desiccation tolerance were subsequently analysed using the Brunch procedure. All contrasts were analysed using a sign test (Purvis & Rambaut 1995). In all phylogenetic analyses, the latest phylogeny available to (sub-) family level from the Angiosperm Phylogeny Group was used (APG II 2003). However, due to the wide range of families and the lack of complete phylogenies to genus level for many families, a series of polytomies were created. Exceptions (classification source given in parentheses) were the Anacardiaceae (Aguilar-Ortigoza & Sosa 2004), Arecaceae (Uhl & Dransfield 1987), Clusiaceae (Gustafsson, Bittrich & Stevens 2002), Fabaceae (Polhill 1994), Lauraceae (Li et al. 2004), and Meliaceae (Muellner et al. 2003).

seed mass and seed-coat ratio

Taxa with desiccation-sensitive seeds had a significantly greater seed mass than taxa with desiccation-tolerant seeds (3383 vs 283 mg, P < 0·05; Table 1, Fig. 1). When correcting for phylogenetic relatedness, this relationship was still significant (P < 0·05; Table 1), indicating that the transition from desiccation-tolerant to desiccation-sensitive seeds is significantly correlated with an increase in seed mass.

The SCR for desiccation-sensitive seeds was significantly lower than for desiccation-tolerant seeds (0·212 vs 0·512, P < 0·05; Table 1, Fig. 1): this relationship was still significant when controlling for both effects of seed mass and phylogenetic relationships (P < 0·05; Table 1).

mean time to germination

In the shade treatment, the desiccation-sensitive seeds germinated more rapidly than desiccation-tolerant seeds (MTG 23·3 vs 47·7 days, P < 0·05; Table 1); the relationship was marginally non-significant in the light treatment (P = 0·061; Table 1). Taking into account the effect of seed mass, germination occurred significantly more rapidly for desiccation-sensitive seeds in both light and shade environments (Table 1). In addition, these relationships were phylogenetically robust (Table 1).

The desiccation-sensitive species were significantly more likely to germinate rapidly in the shade than in the higher-light treatment (sign test, 22 species where MTGS < MTGL; three species where MTGS > MTGL, P < 0·001). However, for the desiccation-tolerant species there was no significant bias to either rapid germination in the shade or higher-light treatments (sign test, 86 species where MTGS < MTGL; 84 species where MTGS > MTGL, P > 0·05).

Across species, there was a highly significant (P < 0·001) relationship between MTG and the spread of germination times (difference between the first and last days on which germination occurred) (Spearman's rank correlation, r_s = 0·70, df = 194; r_s = 0·72, df = 201, for germination in the shade and light, respectively). Thus with an increasing MTG, species had an increasing spread of germination times.

timing of seed dispersal

The distribution of seed dispersal times in relation to the wet and dry seasons was non-random (Table 2; χ² = 11·0, df = 1, P < 0·001), with desiccation-sensitive seeds more likely to be dispersed in the wet than in the dry season (Fig. 2a). However there were exceptions, such as Virola sebeifera for which seed dispersal occurred at the beginning of the dry season (January; Fig. 2a). In contrast, desiccation-tolerant seeds were more likely to be shed in the dry than in the wet season (Fig. 2b).

timing of dispersal and seed desiccation sensitivity

For species with desiccation-sensitive seeds, dispersal occurred predominantly in the wet, as opposed to the dry season, with the pattern reversed for desiccation-tolerant species. Similar results have been demonstrated by Pritchard et al. (2004) for African dryland trees, where the desiccation-sensitive species timed seed dispersal to the wettest month(s) of the year, while desiccation-tolerant seeds were dispersed in either wet or dry months. For desiccation-sensitive seeds, wet-season dispersal has the advantage of minimizing the likelihood of seed desiccation and consequent death. However, for desiccation-tolerant seeds, dry-season dispersal has the advantage of allowing a seasonal seed bank to accumulate prior to the onset of significant rains, when the (pre-)existence of a seed bank may facilitate site pre-emption in advance of species without a seed bank (Garwood 1983; Daws et al. 2005). However, there were exceptions among desiccation-sensitive species. For example, V. sebeifera was dispersed primarily early in the dry season (January). Interestingly, the distribution of this species on BCI is significantly biased towards slope sites (Harms et al. 2001), which maintain a higher level of water availability throughout the dry season. On slope sites, even in the dry season, the matric potential at the soil surface rarely falls below levels that are likely to inhibit germination (approximately −1·5 MPa) (Daws et al. 2002a, 2002b). Consequently, the specialization of this species to wet microsites may circumvent selection for wet-season seed dispersal. This reinforces the hypothesis that desiccation-sensitive seeds are shed to coincide with high water availability, albeit in this case related to spatial rather than temporal patterns.

germination and desiccation sensitivity

In our study the desiccation-sensitive species germinated more rapidly than the desiccation-tolerant species, independently of seed mass and phylogeny. This supports our hypothesis that, based on their high water content (which will minimize the period of imbibition) and metabolic activity at dispersal, desiccation-sensitive seeds will germinate rapidly. Although the observed differences in mean germination times between desiccation-sensitive and desiccation-tolerant species (≈38 days) may seem considerable, imbibition in some large, non-hard-seeded species can be protracted. For example, complete imbibition of dry seeds of Hyophorbe lagenicaulis (Arecaceae) has been reported to take 20 days (Wood & Pritchard 2003). Rapid germination of desiccation-sensitive seeds post-dispersal may enable rapid access to soil water, thereby minimizing the risks of desiccation-induced mortality in short wet-season dry spells.

For desiccation-tolerant species, less rapid germination and a greater spread of germination times may be advantageous. For example, in environments of unpredictable rainfall, such as the start of the wet season on BCI (Garwood 1983), less rapid germination, which is dispersed in time, may reduce the risk of drought or desiccation-induced mortality once seeds have either started to germinate or are at the early seedling stage (Doussi & Thanos 2002). However, this may not be a viable strategy for desiccation-sensitive species, for which slow germination and a prolonged dry spell following seed dispersal could potentially result in mortality of an entire annual cohort of seeds.

For 179 Malaysian tree and shrub species, Foster (1986) reported a positive relationship between seed size and germination rate, which was likely to result from reduced seed–soil contact with increasing seed size. However, the relationship between MTG and desiccation sensitivity that we observed was still significant when controlling for seed mass. While the number of desiccation-sensitive species in the data set used by Foster (1986) is unclear, our findings suggest that the high water content of desiccation-sensitive seeds at shedding and their metabolic activity result in rapid germination, irrespective of seed–soil contact.

In a cross-species analysis, the desiccation-sensitive species were more likely to germinate rapidly in the shade than in the light. In higher-light conditions, there may be a negative impact on large-seeded species as a result of water loss and the consequent decrease in vigour associated with desiccation damage (Pammenter & Berjak 2000). In support of this proposition, Molofsky & Augspurger (1992) reported that in gaps on BCI, seeds of the desiccation-sensitive species Gustavia superba (see Appendix 1) germinate to a higher level when buried under leaf litter than when exposed on the soil surface, presumably because of water loss. However, in the shade, litter had no effect on germination. Similarly, burial can have a beneficial effect on the germination and survival of desiccation-sensitive Quercus rubra seeds (Garcia, Banuelos & Houle 2002). These findings highlight the potential importance of burial by seed predators or burial beneath leaf litter for successful establishment of species with desiccation-sensitive seeds. However, the desiccation-tolerant species were equally likely to geminate rapidly in the shade (lower-shelf) or light (upper-shelf) treatments. For the generally small-seeded desiccation-tolerant species, germination in the higher-light environment may be less inhibited as a result of a greater level of seed–soil contact. Additionally, a number of desiccation-tolerant species have been classified as ‘pioneers’ (sensuSwaine & Whitmore 1988), for example Ochroma pyrimidale and Miconia argentea (Pearson et al. 2002). These species rely on the occurrence of open (high-light) microsites for successful germination and seedling establishment, and have seedlings with a physiological requirement for high-light conditions.

seed mass, seed-coat ratio and desiccation sensitivity

Desiccation-sensitive species in this study had large seeds which will presumably reduce the rate of seed desiccation. Interestingly, our values for average seed mass for desiccation-sensitive and desiccation-tolerant species were similar to values presented by Dickie & Pritchard (2002), who included values from a wide range of both tropical and temperate vegetation types. Our observed trend of desiccation-sensitive seeds being large was phylogenetically robust, leading us to suggest that the loss of desiccation tolerance during evolution has been associated with an increase in seed mass. However large seed size per se does not result in desiccation sensitivity. For example, there are a number of large-seeded desiccation-tolerant species in the data set, including Astrocaryum standleyanum and Dipteryx panamensis, both of which have a seed mass >6 g.

In addition to being metabolically active at seed shed, the thin seed coat of desiccation-sensitive species may also contribute to rapid germination by providing less of a mechanical restraint to germination. Our results for MTG and SCR are consistent with the hypothesis of Pritchard et al. (2004) that rapid germination of desiccation-sensitive seeds may reduce the duration of seed exposure to predation (cf. dipterocarps; Curran & Webb 2000), the corollary being reduced selection for a large investment in seed physical defences. Consequently, per unit mass, desiccation-sensitive seeds appear to be a more efficient use of resources in seed provisioning than desiccation-tolerant seeds.

For some groups of species it has been demonstrated that larger seeds may invest proportionally more resources in defence (e.g. Asteraceae; Fenner 1983). However, the differences in SCR we observed were independent of mass. There is also limited evidence for a negative relationship between seed mass and post-dispersal seed predation (Hulme 1998). Viewed in the light of these two relationships, the generally large-seeded desiccation-sensitive species have a surprisingly low investment in physical defences, which presumably reflects the limited time span of seed exposure to predators. Grubb et al. (1998) demonstrated a positive relationship between both seed physical defences and seed nitrate content for a range of tropical tree seeds. Consequently, it is also possible that a low allocation to defence in desiccation-sensitive species may be related to their being of low nutritional value. Investigations into the chemical composition of desiccation-sensitive and desiccation-tolerant species may be worth pursuing to clarify this issue.