Observational datasets and their utility for predicting genetic differentiation
Our global observational dataset revealed that different combinations of biotic and abiotic factors drove variation on each trait. This trait-specificity would have remained hidden had the environmental and geographical scale of the study been smaller, since we could have not analysed together such a variety of environmental conditions and drivers. In addition, the combination of large-scale field and experimental studies, rarely implemented in evolutionary ecology (but see, e.g., Winn & Gross 1993, Woods et al . 2012), allowed us to assess the potential uses and misuses of observational datasets. In particular, trait-environment relationships inferred from in situpopulations correctly predicted genetic differentiation for reproductive but not vegetative traits. For vegetative traits, the predictability diminished as the presence of plasticity led to interacting or opposing effects of source and exposure environments, as initially forecasted (Fig. 1). The predictability of genetic differentiation was also low for reproductive traits when analysed without accounting for their size-dependence. Therefore, observational data may reliably inform about the current drivers of selection and the adaptive capacity of species only for the traits most closely related with fitness. This might be important for species- and community-level predictive models that rely on trait-environment relationships, and for conservation programs focusing on intraspecific genetic diversity.
Evaluating trait-environment relationships can also be useful for predicting plant performance in populations introduced outside native ranges (Alexander et al . 2012, Hulme & Barrett 2013). InP. lanceolata , traits showed broadly similar correlations with environmental factors in both native and non-native ranges, in agreement with previous work in other taxa (Maron et al . 2004, Montagueet al . 2008, Rosche et al . 2019; but see Keller et al . 2009). Notably, the similarities in trait patterns between ranges held despite the location of non-native populations in warmer and more arid conditions. This suggests that the trait-environment correlations largely persist for some species even if they occupy more extreme areas of environmental space, facilitating ecological predictions in a context of global change. Yet some trait-environment correlations observed inP. lanceolata were weaker in the non-native range (see also Alexander et al . 2012). This finding highlights that genetic differentiation may be less predictable for non-native populations and that a total equivalence in trait patterns between ranges cannot be taken for granted due to potential evolutionary divergence. The presence of weaker trait-environment relationships in non-native populations may be due to a higher role of plasticity (although the latter is not clearly supported by a recent meta-analyses across species; see Palacio-López & Gianoli 2011), or may instead result from repeated introductions in the non-native range (Smith et al . 2020). Further sudies on widespread species might help to clarify the processes and patterns resulting from ecological and evolutionary divergence at large spatial scales. In particular, our observational network can form the basis for future experimental work.