INTRODUCTION
Large vertebrates are being extirpated across the tropics, which affects the myriad tree species that interact with these animals (Kurten 2013; Dirzo et al. 2014). Such defaunation can affect plants in many ways. When trees are left without seed dispersers, for example, they may suffer population declines (Brodie et al. 2009; Culot et al. 2017; Rogers et al. 2017). Because many trees dispersed by large vertebrates are themselves large or have dense wood, defaunation may even induce shifts in tree species composition that reduce the aboveground biomass of tropical forests, with implications for the global carbon cycle (Brodie & Gibbs 2009; Bello et al. 2015; Osuri et al. 2016; Peres et al. 2016). Furthermore, if hunting removes large predators, granivore populations could increase, leading to reduced seed survival (Galetti et al. 2015; Rosin & Poulsen 2016). However, many tropical trees experience very strong conspecific density dependence (Harms et al. 2000; Peters 2003; Comita et al. 2010; Terborgh 2012), which implies that lower survival at early stages (e.g. through reduced seed dispersal or enhanced seed predation) could potentially be offset at the population level by ameliorated density dependence. Moreover, many of the hunted vertebrates are potent seed predators (Roldán & Simonetti 2001; Donattiet al. 2009) or trample seedlings (Rosin et al. 2017), so removing these animals could benefit regeneration in certain plant species. It is critical, therefore, to assess how defaunation affects not just seed dispersal or seedling survival, but the entire life cycle of tropical trees.
Most previous studies on how defaunation affects trees (particularly those focusing on forest carbon impacts) have focused almost exclusively on reduced seed dispersal. These studies often simulate community composition in defaunated forests by ‘removing’ tree species that are large vertebrate-dispersed (Peres et al. 2016; Chanthorn et al. 2019) or that have large seeds (Bello et al. 2015; Osuriet al. 2016), and show that this can result in substantial reductions in aboveground biomass (i.e. carbon storage). Looking at empirical evidence from defaunated forests, though, the patterns are less clear. Populations of a tree species that significantly contributed to carbon stocks were indeed declining in defaunated forests in the Brazilian Atlantic Forest (Culot et al. 2017), but hunting-induced dispersal limitation appeared to have no impact on total tree biomass in Malaysian Borneo (Harrison et al. 2013).
The best way to predict the effects of defaunation on tree species is to conduct population-level, whole-life-cycle analyses. But such analyses are very resource-intensive, precluding the evaluation of all tropical tree species over any conservation-relevant time frame. Therefore, it is important to try to ascertain whether we can predict a prioriwhich tree species might be susceptible to defaunation, for example based on their phenotypic traits. Defaunation responses could potentially vary with morphological traits such as seed size, which may affect seed predation (Mendoza & Dirzo 2007), or with ecological traits such as dispersal mode, which could affect susceptibility to disperser loss (Peres et al. 2016). However, given the multiple effects of defaunation on plants at different life stages, what matters is how all of the impacts combine to influence overall population dynamics (Harrison et al. 2013) and whether this varies with life history characteristics. Demographic rates for tropical trees tend to be correlated with physical traits (Poorter et al. 2008) that are easier to measure and collect. If we could use combinations of physical traits to ascertain a given tree species’ susceptibility to defaunation-induced population decline, we could better predict the changing species composition of defaunated tropical forests.
Here we synthesized data on tropical tree populations, the multiple impacts of defaunation across the plant life cycle, and tree morphological and demographic traits to assess whether we can predict how trees with different traits vary in their responses to hunting-induced population declines. We used density-dependent demographic models and Monte Carlo simulations that incorporate data on all facets of tree life history, defaunation effects, and the (often substantial) uncertainty in these factors. Specifically, our objective was to determine whether any traits or trait combinations were associated with tree susceptibility to a range of different defaunation effects.