Aim: Our understanding of the mechanisms that maintain forest diversity under changing climate can benefit from knowledge about traits that are closely linked to fitness. We tested whether the link between traits and seed number and seed size is consistent with two hypotheses, termed the leaf economics spectrum and the plant size syndrome, or whether reproduction represents an independent dimension related to a seed size–seed number trade-off.

Location: Most of the data come from Europe, North and Central America and East Asia. A minority of the data come from South America, Africa and Australia.

Time period: 1960– 2022.

Major taxa studied: Trees.

Methods: We gathered 12 million observations of the number of seeds produced in 784 tree species. We estimated the number of seeds produced by individual trees and scaled it up to the species level. Next, we used principal components analysis and generalized joint attribute modelling (GJAM) to map seed number and size on the tree traits spectrum.

Results: Incorporating seed size and number into trait analysis while controlling for envi-ronment and phylogeny with GJAM exposes relationships in trees that might otherwise remain hidden. Production of the large total biomass of seeds [product of seed number and seed size; hereafter, species seed productivity (SSP)] is associated with high leaf area, low foliar nitrogen, low specific leaf area (SLA) and dense wood. Production of high seed numbers is associated with small seeds produced by nutrient-demanding species with softwood, small leaves and high SLA. Trait covariation is consistent with opposing strategies: one fast-growing, early successional, with high dispersal, and the other slow-growing, stress-tolerant, that recruit in shaded conditions.

Main conclusions: Earth system models currently assume that reproductive allocation is indifferent among plant functional types. Easily measurable seed size is a strong predictor of the seed number and species seed productivity. The connection of SSP with the functional traits can form the first basis of improved fecundity prediction across global forests.

Lack of tree fecundity data across climatic gradients precludes the analysis of how seed supply contributes to global variation in forest regeneration and biotic interactions responsible for biodiversity. A global synthesis of raw seed production data shows a 250-fold increase in seed abundance from cold-dry to warm-wet climates, driven primarily by a 100-fold increase in seed production for a given tree size. The modest (threefold) increase in forest productivity across the same climate gradient cannot explain the magnitudes of these trends. The increase in seeds per tree can arise from adaptive evolution driven by intense species interactions or from the direct effects of a warm, moist climate on tree fecundity. Either way, the massive differences in seed supply ramify through food webs potentially explaining a disproportionate role for species interactions in the wet tropics.

Alternative stable state theory predicts that diferent disturbance regimes may support contrasting ecosystem states under otherwise analogous environmental conditions. In fre-prone systems, this theory is often invoked to explain abrupt ecotones, especially when adjacent vegetation types have contrasting fammabilities and difering tolerances to pyric perturbation. Despite being well-documented in forest-savanna transitions, unambiguous examples of fre-driven alternate stable states (FDASS) in arid systems are rare. The current study examined whether fammable spinifex (Triodia spp.) grasslands and fre-sensitive waputi (Aluta maisonneuvei subsp. maisonneuvei) shrublands in Australia’s Gibson Desert represent FDASS. Specifcally, analyses of soil and topographic variables assessed whether environmental diferences explain habitat zonation. To determine whether diferent fammabilities of Aluta and Triodia systems may perpetuate alternative states via vegetationfre feedback processes, community-level fuelloads were quantifed to provide an indirect measure of fammability. To determine the propensity for fre to trigger ‘state-shifting’, community responses to a single high-severity fre were evaluated. Habitat segregation did not relate to between-site environmental diferences, and the fuel-load study indicated that the more pyrophytic community (Triodia grassland) is more fammable, and hence more likely to experience higher frequency frecycles. Fire was identifed as a potential vector of ‘state-change’, because although both systems regenerated well after fre, Triodia reproduced more prolifcally at a younger age than Aluta, and hence should tolerate shorter fre-return intervals. In the absence of between-community topographic and edaphic diferences, or herbivores that consume either plant, it is likely that Aluta shrublands and Triodia grasslands represent fre-mediated alternative equilibrium states.

Large, high-severity wildfires, or "megafires," occur periodically in arid Australian spinifex (Triodia spp.) grasslands after high rainfall periods that trigger fuel accumulation. Proponents of the patch-burn mosaic (PBM) hypothesis suggest that these fires are unprecedented in the modern era and were formerly constrained by Aboriginal patch burning that kept landscape fuel levels low. This assumption deserves scrutiny, as evidence from fire-prone systems globally indicates that weather factors are the primary determinant behind megafire incidence, and that fuel management does not mitigate such fires during periods of climatic extreme. We reviewed explorer's diaries, anthropologist's reports, and remotely sensed data from the Australian Western Desert for evidence of large rainfall-linked fires during the pre-contact period when traditional Aboriginal patch burning was still being practiced. We used only observations that contained empiric estimates of fire sizes. Concurrently, we employed remote rainfall data and the Oceanic Niño Index to relate fire size to likely seasonal conditions at the time the observations were made. Numerous records were found of small fires during periods of average and below-average rainfall conditions, but no evidence of large-scale fires during these times. By contrast, there was strong evidence of large-scale wildfires during a high-rainfall period in the early 1870s, some of which are estimated to have burnt areas up to 700,000 ha. Our literature review also identified several Western Desert Aboriginal mythologies that refer to large-scale conflagrations. As oral traditions sometimes corroborate historic events, these myths may add further evidence that large fires are an inherent feature of spinifex grassland fire regimes. Overall, the results suggest that, contrary to predictions of the PBM hypothesis, traditional Aboriginal burning did not modulate spinifex fire size during periods of extreme-high arid zone rainfall. The mechanism behind this is that plant assemblages in seral spinifex vegetation comprise highly flammable non-spinifex tussock grasses that rapidly accumulate high fuel loads under favorable precipitation conditions. Our finding that fuel management does not prevent megafires under extreme conditions in arid Australia has parallels with the primacy of climatic factors as drivers of megafires in the forests of temperate Australia.