Conclusion
Habitat loss altered the effects of macroclimate on community assembly, calling attention to the scale-dependence of trait-mediated composition in human-modified landscapes. Relatively weak ability of individual traits to predict abundance and opposing axes of trait-mediated assembly suggest the need to examine trait-coordination for multiple stressors (Craine et al., 2012; Shen et al., 2019). Moreover, ‘soft’ traits measured here clearly do not provide sufficient insights into community-level responses to climate gradients (Li et al., 2015; Shipley et al., 2016). Measuring a wider suite of traits relevant to different stressors will help capture the full spectrum of functions that govern compositional change in human-modified forests (Maréchaux et al., 2019; Medeiros et al., 2019; Shipley et al., 2016). For example, no studies have yet correlated hydraulic traits to species performance in the warmer, drier microclimates near fragment edges. Shifts in the success of phenotypes with habitat loss could be due to edge effects that alter forest microclimates (Arroyo-Rodríguez et al., 2017; Davis et al., 2019): e.g., warmer, drier conditions near edges may amplify water stress during the dry season (Ewers and Banks-Leite, 2013). Examining how microclimate mediates the impacts of macroclimate on species performance would help predict compositional change in human-modified landscapes subject to a changing climate (De Frenne et al., 2019; Zellweger et al., 2020). Hierarchical, scale-dependent insights into trait-environment linkages will allow evidence-based restoration planning to achieve different targets such as community structure, diversity or ecosystem functioning (Laughlin, 2014; Laughlin et al., 2017).