DISCUSSION
Overhunting can alter many plant-animal interactions, with consequences ranging from altered dynamics of tropical tree populations (Brodieet al. 2009; Culot et al. 2017) to potential declines in forest carbon storage (Bello et al. 2015; Peres et al.2016; Chanthorn et al. 2019). But we still have a limited understanding of how the multiple effects of defaunation combine to affect tree populations. Our results suggest that how a tropical tree species responds to defaunation is primarily based on whether and how the predation of its seeds is altered by hunting. But unfortunately, even knowledge about altered seed predation rates still confers only limited ability to predict population-level responses. Moreover, our knowledge about how defaunation affects seed predation remains limited. Granivory could increase (Galetti et al. 2015; Rosin & Poulsen 2016; Culot et al. 2017) or decrease (Wright et al. 2000; Roldán & Simonetti 2001; Beckman & Muller-Landau 2007) under defaunation and, even in the same forest, species can differ wildly in how seed predation rates change in response to defaunation (Guariguataet al. 2000; Rosin & Poulsen 2016). In our model, we applied changes in seed predation equally to all of our species, but our results might have changed if seed predation varied according to species traits. For example, Mendoza and Dirzo (2007) hypothesized that smaller seeds, whose small-bodied predators would likely remain extant even in highly defaunated systems, would face higher predation pressure than larger seeds, whose predators would be eliminated by overhunting. However, we conducted a review of studies that measured seed predation in defaunation versus non-defaunation conditions (Table S4) and found that seed size did not predict changes in predation pressure (Figure S4). Though seed size did not correlate with change in seed predation, our results show that small-seeded angiosperms were more demographically sensitive to changes in seed predation than gymnosperms or large-seeded species. This result provides a potential focus for future research: understanding exactly how defaunation affects seed predation may be particularly important for assessing the vulnerability of small-seeded angiosperms.
Defaunation can also affect later life-stages via changes in herbivory or other physical damage (Gardner et al. 2019). We did not include such effects in our model because of very scant data connecting defaunation to changes in sapling or adult vital rates. Removing large-bodied ecosystem engineers such as elephants may impact older plant life stages (Poulsen et al. 2018), though the distribution of such megafauna and their foraging behaviour are highly variable across the world’s tropical forests. While changes in herbivores and ecosystem engineer abundances may significantly impact local plant communities (Luskin et al. 2017), it is difficult to consider these factors in a global assessment of defaunation.
Surprisingly, factors such as dispersal mode and the strength of density dependence were unrelated to demographic responses to defaunation in our model. Previous studies have assumed that dispersal limitation was the most important effect of defaunation (Terborgh 2013), leading to predictions that defaunation-induced losses of seed dispersal will reduce carbon storage (Brodie & Gibbs 2009; Bello et al. 2015; Dantas de Paula et al. 2018; Chanthorn et al. 2019). However, given that defaunation alters seed predation as well as seed dispersal, and that altered seed dispersal was unrelated to population-level defaunation impacts, overhunting may not necessarily cause the widespread replacement of heavy-wooded, vertebrate-dispersed species by lighter-wooded species (with other means of dispersal) that have been predicted. Therefore, whether defaunation will result in major losses of forest carbon remains unclear. Indeed, not all defaunated forests show decreases in biomass (Harrison et al. 2013; Bagchiet al. 2018). Despite attempts to model the impacts of defaunation on tropical tree communities and ecosystems via altered seed dispersal alone, our study shows that there is too much uncertainty surrounding other defaunation effects—notably seed predation—to currently be able to make accurate predictions about cascading impacts on populations and ecosystem processes.
Likewise, phenotypic traits conferred little ability to predict defaunation responses across species. While many of the traits did correlate with the vital rates that determine population dynamics, these correlations were relatively weak (Table S6). Importantly, of the traits that we assessed, only seed mass significantly correlated with adult survival, which, in long-lived woody plants, influences population dynamics more than any other vital rate (Silvertown et al. 1993; Zuidema 2000). Because large seeds are unlikely to be dispersed by small vertebrates, many studies have assumed that large-seeded species are the most vulnerable to defaunation (e.g. Bello et al. 2015; Osuriet al. 2016). However, in our study, large-seeded species were less responsive to defaunation because of their lower adult mortality. Our smaller-seeded species had stronger responses to defaunation, but the direction of the response was mediated by how defaunation affected seed predation. As seed mass does not predict changes in seed predation (Figure S4), we still cannot use this trait to assess a priorispecies’ responses to defaunation.
What, then, can we predict about the consequences of defaunation for tropical trees? We know from studies of individual species that defaunation can cause significant population declines (Wright & Duber 2001; Culot et al. 2017), but we still have limited ability to scale up these observations to predict patterns at the community or ecosystem level. Phenotypic traits and even demographic vital rates are not sufficient to predict defaunation effects. However, our results highlight a fruitful direction for future research; an improved understanding of seed predation—how it affects tree populations and how it is affected by defaunation—would greatly reduce uncertainty around the cascading impacts of defaunation. This will improve the accuracy of predictions of the global consequences of overhunting for tropical forests.