Introduction
Functional traits are morphological, physiological or phenological features of organisms that influence the components of fitness, i.e. survival and reproduction (Reich et al . 2003, Violle et al . 2007, Adler et al . 2014). Intraspecific variation in functional traits is widely documented and has important implications for population dynamics (Hughes et al . 2008, Villellas & García 2017), evolutionary trajectories (Moran et el. 2016, Caruso et al. 2020), community assembly (Violle et al . 2012, Des Rocheset al . 2018), and ecosystem functioning (Crutsinger et al . 2006, Breza et al . 2012). Disentangling the environmental drivers of functional trait variation is thus of great ecological and evolutionary interest (Liancourt et al. 2013, van de Pol et al 2016, Bruelheide et al 2018), and can improve predictions of species responses to global change (Benito Garzón et al . 2011, Violle et al . 2014, Moran et al . 2016).
The predominant approach to identify the drivers of functional trait variation has relied upon assembling large databases of observedin situ trait variation (e.g., Enquist et al . 2016, Moranet al . 2016, Iversen et al . 2017, Kattge et al . 2020) and the association of these trait values with candidate environmental drivers. However, interpreting trait-environment relationships inferred from observational field datasets requires understanding the processes underlying trait variation. Intraspecific trait variation observed in situ among populations may arise from genetic differentiation and/or phenotypic plasticity (Chevin et al . 2010, Merilä & Hendry 2014). Across large environmental gradients, genetic differentiation among populations can result from adaptation to local conditions (but see the role of neutral and historical processes in Keller et al . (2009), Santagelo et al . (2018)). Genetically determined traits are thus expected to show correlations with the source environment. However, genetic differentiation might be obscured by phenotypic plasticity (which can also be adaptive; see Matesanz et al . 2010, Palacio-López et al. 2015), reducing the consistency of trait-environment relationships across environmental contexts.
Combining experimental and in situ field data enables us to assess the potential uses and misuses of observational trait datasets. A common way to partition trait variation is through a common garden experiment (Clausen et al. 1940, MacColl 2011, Franks et al . 2014). Specifically, by growing offspring from multiple provenances together in a set of controlled conditions, we can disentangle the effects of source environments (leading to genetic differentiation) from those of exposure environments (driving phenotypic plasticity). Notably, by evaluating the different scenarios involving genetic and plastic effects on traits, we can assess the utility of observational data for predicting genetic differentiation (Fig. 1). For example, a predominance of genetic over plastic effects decreases the relative importance of genotype-by-environment interactions, and increases the predictability of trait values from average environmental conditions of source populations (Fig. 1a,f). In contrast, a high level of plasticity causes traits to be more strongly determined by the exposure environment, decreasing trait predictability from source environment (Fig. 1b,g,h). Source and exposure environments can have similar or opposing effects on traits (Fig. 1c-e), with opposing effects known as countergradient variation (Conover & Schultz 1995, Conover et al . 2009). Countergradient variation may lead to an apparent absence of trait variation among populations in the field (Fig. 1d), or even to patterns counter to those of genetic differentiation (Fig. 1e).
The role of genetic differentiation and phenotypic plasticity on intraspecific variation differs among functional traits (Albert et al . 2010a, Funk et al . 2017, Münzbergová et al . 2017). Species may show evolutionary conservation of the traits most directly related to fitness through genetic differentiation (Scheiner 1993, Stearns & Kawecki 1994, Sih 2004), and instead display plasticity in underlying morphological or physiological traits, to buffer environmental perturbations (Sultan 1995, Richards et al . 2006). This view is also supported by demographic studies finding that the most influential processes on population growth rate show relatively low variability (Pfister 1998, Burns et al . 2010, Hilde et al . 2020; but see McDonald et al . 2017). In plants, vegetative traits often show higher plasticity than reproductive traits (Bradshaw 1965, Schlichting & Levin 1984, Frazee & Marquis 1994). For example, both biomass and reproductive investment per unit biomass determine reproduction, but while biomass is expected to show high plasticity due to its influence on several demographic parameters (Harper 1977), reproductive investment per unit biomass may be more conserved. This might be especially true for short-lived taxa, in which reproduction usually has the highest influence on population growth (Silvertownet al . 1996, García et al . 2008, Shefferson and Roach 2012). Yet reproductive investment may appear to be strongly driven by plasticity if evaluated at the whole plant level, due to the inclusion of a more labile biomass-dependent component (Biere 1995, Weineret al . 2009). It is important therefore to partition reproductive traits into biomass dependent and independent components, to better understand the role of genetic differentiation and plasticity.
Despite the abundance of studies analysing trait-environment relationships at local or regional scales (e.g., Oleksyn et al . 1998, Villellas & García 2013, Preite et al . 2015, Münzbergováet al . 2017), there is a critical gap in knowledge about the drivers of intraspecific trait variation at global scales (MacColl 2011). Environmental effects may be difficult to detect if drivers are assessed independently from each other, or if studies omit significant parts of a species’ environmental niche (Matesanz et al . 2010, Hulme & Barrett 2013, Shipley et al. 2016). Widespread plants offer a unique opportunity to unravel the multiple drivers of trait variation from local to global scales. While some studies have analysed trait genetic differentiation and plasticity across species’ ranges (e.g., Joshi et al . 2001, Maron et al . 2004, Alexanderet al . 2012), we lack global assessments of the responses of different types of traits to multiple environmental drivers using the combined power of experimental and observational data.
Here we analyse responses of vegetative vs. reproductive traits of the short-lived herb Plantago lanceolata to a set of environmental drivers, both in a common garden and in the field. By growing individuals from multiple populations under several light and water conditions, we tested 1) whether vegetative traits (plant biomass, specific leaf area and root:shoot ratio) showed higher levels of plasticity than reproductive traits (probability of flowering and fecundity), and 2) whether reproductive traits showed more consistent population genetic differentiation across exposure treatments than vegetative traits, and higher consistency between genetic and plastic responses. To account for the potential size-dependency of plant reproductive investment, we examined reproductive traits by both including and excluding plant biomass as a covariate in the analyses. Finally, by comparing experimental results with trait-environment relationships detected from a global-scale observational survey, we evaluated 3) whether observational data provided a better prediction of genetic differentiation for reproductive than vegetative traits.