Limits to reproduction and seed size-number trade-offs that shape forest dominance and future recovery

Abstract

The relationships that control seed production in trees are fundamental to understanding the evolution of forest species and their capacity to recover from increasing losses to drought, fire, and harvest. A synthesis of fecundity data from 714 species worldwide allowed us to examine hypotheses that are central to quantifying reproduction, a foundation for assessing fitness in forest trees. Four major findings emerged. First, seed production is not constrained by a strict trade-off between seed size and numbers. Instead, seed numbers vary over ten orders of magnitude, with species that invest in large seeds producing more seeds than expected from the 1:1 trade-off. Second, gymnosperms have lower seed production than angiosperms, potentially due to their extra investments in protective woody cones. Third, nutrient-demanding species, indicated by high foliar phosphorus concentrations, have low seed production. Finally, sensitivity of individual species to soil fertility varies widely, limiting the response of community seed production to fertility gradients. In combination, these findings can inform models of forest response that need to incorporate reproductive potential.

Fig. 1. Seed production quantifies forest regeneration potential.

Regeneration of forests devastated by multi-year drought and fire depend on a vastly diminished seed supply. a Seed production is limited to unburned landscape fragments in the Sierra Nevada mixed conifer zone following 2020 burns at a Masting Inference and Forecasting network (MASTIF) and National Ecological Observatory Network (NEON) site (Shaver Lake, CA). b Total reproduction includes not only seeds, but also defenses, including wood, spines, and resin flow in conifer cones; examples from the heavily burned Sierra Nevada and Coast ranges include Calocedrus decurrens, Pinus albicaulis, P. contorta, P. coulteri, P. flexilis, P. lambertiana, P. monophylla, P. monticola, P. ponderosa, P. radiata, P. sabiniana, Pseudotsuga menzesii, Sequoiadendron giganteum, and Tsuga mertensiana. Mass fractions for seeds to seeds plus cones ranges from 3% for P. radiata, P. contorta, P. coulteri, and P. sabiniana to 61% for C. decurrens. The largest cone in b (Pinus lambertiana) is 46 cm. Photo credits: James S. Clark and Jordan Luongo.

Fig. 2. seed size-number trade-offs and species seed production.

a Species seed production (SSP, g seed per m² tree basal area) is not constrained by the strict size-number trade-off (dashed line with a slope of zero). Instead, it varies over ten orders of magnitude and has a positive correlation with seed mass across 714 tree species (${\log }_{10}$SSP $=4.29+0.546\cdot {\log }_{10}m$, R² = 0.189, p < 10⁻¹⁵, n = 714). b SSP exhibits phylogenetic coherence for 482 species having phylogeny data (68% of species). Brown and green text highlight species that produce coniferous cones and fleshy fruits, respectively. The phylogenetic signal is estimated using Pagel's λ = 0.60 (p < 10⁻⁹, n = 482).

Fig. 3. Effects of foliar nutrients on seed production.

Effects of foliar nitrogen (N) and phosphorus (P) on SSP (g seed per m² tree basal area) from the model in Supplementary Table 1 plotted for the broadleaf deciduous leaf habit (other leaf types exhibit same patterns). The convex hull for the surface is restricted to the data coverage. Symbols indicate leaf habit, including broadleaf deciduous (BD), broadleaf evergreen (BE), needleleaf evergreen (NE), and scalelike evergreen (SE).

Fig. 4. Effects of soil fertility on seed production.

Sensitivity of individual standardized production (ISP) to soil fertility based on within-species response to cation exchange capacity (β_cec). Text color follows Fig. 2. Red dashed line indicates the ancestry of gymnosperms. Percentages of species that respond negatively to CEC are labeled for species groups. The analysis includes 141 species that span a sufficient CEC range in the Masting Inference and Forecasting (MASTIF) network to estimate a robust effect. The phylogenetic signal, estimated for 129 species (91% of species) having phylogeny data, is highly significant (Pagel's λ = 0.87, p < 0.001, n = 129).

Fig. 5. The MASTIF model, summarized from Clark et al., includes three levels, observed responses (above), process, and parameters (part of the posterior distribution, middle), and predictors (below).

The data model for observed responses includes uncertainty that comes from seed dispersal (seed traps), the fraction of the crop that can be observed (crop counts), and detection of mature status. The process model describes change in maturation status ρ_i,t and, once mature, conditional fecundity ψ_i,t. Fitted coefficients for conditional fecundity β^ψ and maturation probability β^ρ describe how predictor variables in red affect maturation and fecundity. Error in the process model is absorbed by process error variance σ². Predictors are held in a design matrix x_i,t for conditional fecundity. Diameter d_i,t is the predictor for maturation status. Additional subscripts for location j and species s in the main texts are suppressed to reduce clutter.