Predictors of intraspecific trait variation in Puerto Rican forests: Bayesian modeling approach and causal understanding of drivers

Background/Question/Methods

: There is significant evidence of a large degree of intraspecific variation in plant traits. This intraspecific trait variation (ITV) has been hypothesized to lead to plant coexistence. Specifically, theoretical work suggests that coexistence will be promoted if ITV induces a mean-variance tradeoff, that where fitness is highest for a species, there is less variability. We test this hypothesis using data for ITV in wood density and specific leaf area (SLA) for 29 tree species across 24 forest plots in Puerto Rico. Species abundance and basal area were measured in plots along precipitation (900-3170 mm/yr), soil (limestone and volcanic), and elevation gradients (5-985 meters). We derived site favorability using herbaria data and Maxent species distribution models. First, we test this hypothesis using a series of linear models comparing site favorability to magnitude of ITV. We also fit linear models testing how site rainfall and topographic variability relate to ITV. Then, we fit a hierarchical model to understand how site characteristics (such as environmental variability), species occurrence, and site favorability lead to ITV in a species at a site.

Results/Conclusions

: Species site-specific mean trait values are correlated with environmental variables. Wood density positively responds to precipitation (β=0.031 ± 0.014) and negatively to elevation (β=-0.022±0.008). ITV in wood density is only an order of magnitude lower than interspecific variation (variance of 0.081 versus 0.153), though there is much more inter- than intraspecific variation for specific leaf area (variance of 1252.477 m2.kg−1 versus 10720.2 m2.kg−1). Preliminary results show site and species are significant predictors of ITV in both wood density and SLA (p< 0.005 in an ANOVA).Most study species (18/29) followed our prediction with a negative relationship between ITV in wood density and habitat suitability. In contrast, 18/29 of species had positive relationships between ITV in SLA with habitat suitability. Site terrain roughness positively correlated with ITV of wood density in most cases (20/27), but this pattern did not hold for ITV of SLA (13/30 slopes were positive). Lastly, ITV was higher in species that occur across a broader range of precipitation and elevation. Taken together, these results indicate mixed support for the hypothesis that ITV declines as site favorability increases. We are currently modeling the magnitude of ITV in a hierarchical Bayesian model to incorporate these multiple causal factors.