The cost of fruit and the penalty of youth: Predicting mean annual seed production in single-species forest stands

Abstract

Quantifying mean annual tree seed production is important for conservation and forestry applications, but its estimation remains a substantial challenge. Interspecies variation in seed production is often expressed as a trade-off between seed size and seed number, forming a key component of established models of mean annual seed production in forest stands. In this study we improved on these models by quantifying additional effects from accessory costs and reproductive maturity status, using data from temperate-zone Eucalyptus species to assess model performance.Firstly, by evaluating a range of reproductive traits, we found that seed rain density was more closely associated with ‘fruit cost per seed’ measures than seed size measures. For the most common measure of seed rain density (based on direct observations of free seed fall), woody capsule surface area per viable seed explained 70% of observed variation, while seed mass explained only 28% (n = 16 species).Secondly, an existing model based on seed mass, tree size and tree density was extended to include a stand maturity function, and to use alternative reproductive traits. For a smaller dataset with long-term observations and detailed stand measurements, the proportion of variance explained by the extended model ranged from 81 to 98% depending on trait selection (mean = 93%), with published models explaining less variance (75–83%, mean = 80%).By reframing interspecies variation in seed production as a trade-off between seed cost and seed number, we found evidence that investment in heavy fruit structures may represent a substantial constraint on seed output among forest stands from a common genus and region, although our results must be considered somewhat preliminary, given the small number of species examined. We have also developed an enhanced predictive model, which could be used with stand-level models for tree size and tree density to improve predictions of mean annual forest seed production in landscape simulations.