Seed storage in woody fruits on plants has been much studied, whereas trait variation in seed release has been given scant attention. In non-Mediterranean climates, some species release seeds immediately after fire, whereas others retain seeds in open fruits/cones for longer. We expected that species with wind-dispersed seeds and those killed by fire would spread their recruitment risks by having stronger cues for fruit opening and slower seed release once fruits were open. We therefore tested whether fire intensity (heat) affected fruit opening and seed release in 19 species. We then contrasted fruit opening and seed release among (1) serotiny levels (weak, moderate, strong), (2) dispersal (wind v. unassisted) and (3) resprouting ability (killed v. resprout) traits. Only three species required heat for fruit opening. Most species, however, retained varying proportions of seeds in open fruits. Strongly and moderately serotinous species retained seeds in open fruits longer than did weakly serotinous species. Both species with wind-dispersed seeds and fire-killed species required stronger heat effects for fruits to open but retained seeds in open fruits longer than did species with alternative traits. By delaying seed release after fruits have opened, species with wind-dispersed seeds, and those that are killed by fire, maximise the ability of seeds to arrive at safe sites after fire.

Delayed seed release (serotiny) is a convergent plant trait in fire-prone regions of the world but explaining the degree of serotiny has remained elusive because of the paucity of community data. Selective forces involving seed predators, fire and soil nutrients have been suggested as factors influencing serotiny. We tested whether protection of seeds and/or synchronized dispersal were associated with different levels of serotiny and if resprouting ability influences selection for strong serotiny. We compared the numbers and abundance of 146 woody species with delayed dispersal among five community types varying in combinations of fire severity, fire frequency, soil fertility and seed predators. The strength of the relationship between levels of serotiny and environmental factors was tested among community types ranging from rainforests to heathlands. Highest levels of serotiny were recorded in low nutrient shrublands with intermediate fire return intervals that burn at high severity, while the lowest were recorded in high nutrient, low flammability forests. Both protection of seeds and synchronized seed release were related to fire effects in nutrient-limited environments. Strong serotiny is prominent in species killed by fire whereas weak serotiny is more common in resprouting species. Recruitment failure in the inter-fire interval appears to drive selection for strong maternal care of seeds and synchronized seed dispersal in fire-prone environments. Weak serotiny is proposed as a bet-hedging strategy that relies on resprouting after fire for population persistence and higher probability of inter-fire recruitment. The spectrum of serotiny (weak to strong) in these communities is proposed to be driven by the interactive effect of both fire and soil nutrients on the selection for delayed seed dispersal.