Community‐level natural selection modes: A quadratic framework to link multiple functional traits with competitive ability

Abstract

Research linking functional traits to competitive ability of coexisting species has largely relied on rectilinear correlations, yielding inconsistent results. Based on concepts borrowed from natural selection theory, we propose that trait–competition relationships can generally correspond to three univariate selection modes: directional (a rectilinear relationship), stabilising (an n‐shaped relationship), and disruptive (a u‐shaped relationship). Moreover, correlational selection occurs when two traits interact in determining competitive ability and lead to an optimum trait combination (i.e., a bivariate nonlinear selection mode). We tested our ideas using two independent datasets, each one characterising a group of species according to (a) their competitive effect on a target species in a pot experiment and (b) species‐level values of well‐known functional traits extracted from existing databases. The first dataset comprised 10 annual plant species frequent in a summer‐rainfall desert in Argentina, while the second consisted of 37 herbaceous species from cool temperate vegetation types in Canada. Both experiments had a replacement design where the identity of neighbours was manipulated holding total plant density in pots constant. We modelled the competitive ability of neighbours (i.e., the log inverse of target plant biomass) as a function of traits using normal multiple regression. Leaf dry matter content (LDMC) was consistently subjected to negative directional selection in both experiments as well as to stabilising selection among temperate species. Leaf size was subjected to stabilising selection among desert species while among temperate species, leaf size underwent correlational selection in combination with specific leaf area (SLA): selection on SLA was negative directional for large‐leaved species, while it was slightly positive for small‐leaved species. Synthesis. Multiple quadratic regression adds functional flexibility to trait‐based community ecology while providing a standardised basis for comparison among traits and environments. Our analyses of two datasets from contrasting environmental conditions indicate (a) that leaf dry matter content can capture an important component of plant competitive ability not accounted for by widely used competitive traits, such as specific leaf area, leaf size, and plant height and (b) that optimum relationships (either univariate or bivariate) between competitive ability and plant traits may be more common than previously realised.

Inferring trait‐based mechanisms and variance consequences of abiotic secondary dispersal from the spatio‐temporal distribution of a desert soil seed bank

“…Qualitatively similar results were found in the Sonoran Desert in North America, where abiotic factors are thought to play an important role in seed redistribution (Reichman, 1984; but see Venable et al, 2008). Debris microsites thus offer large amounts of seeds that have been after-ripened under natural field conditions and can thus be readily used in plant experiments (Rolhauser et al, 2019) and in restoration experiences. Importantly, seeds of many abundant annuals that are highly associated with shrubs (such as Bowlesia incana, Schismus barbatus, Chenopodium papulosum, Descurainia erodiifolia and Amaranthus standleyanus; Rolhauser & Pucheta, 2016) were highly abundant in debris samples (Appendix S2).…”

“…We used quadratic mixed logistic regression models to describe the relationship between each of the response variables X i and the combined effects of seed mass and shape (see Figure 1). Quadratic terms provide models with flexibility that may be important in describing cross-species trait functionality (Rolhauser & Pucheta, 2017;Rolhauser et al, 2019). Each model (one for each donor microsite) had the following general form where π i is the expected proportion of species i in the jk cohort and plot combination, and ln i 1− i is the logit function that links π i to its linear predictor (Gelman & Hill, 2006;Faraway, 2016).…”

“…However, seed traits have not been used as quantitative predictors of seed spatial distribution. Recently, trait-based studies have emphasized the importance of testing interactions between traits, since the effect of one trait may depend on the values of another (Laughlin & Messier, 2015;Yang et al, 2018;Rolhauser et al, 2019). Our first general goal here is to quantify interactive effects of seed mass and shape on seed's "microhabitat choice", i.e., on among-microsite seed bank distribution (and this goal leads to question 2 below).…”

Getting to the root of restoration: considering root traits for improved restoration outcomes under drought and competition

“…Although high SRL may improve a plant's competitive effect, influences on performance measures may be context specific. For example, Rolhauser et al (2019) found that high SRL was negatively related to competitive effect, likely due to interactions with pulse moisture regimes. Root length (Gordon & Rice 1993; Leger & Goergen 2017) and root area (Wang et al 2010), both traits related to resource acquisition, have also been linked to species competitive effects but results are limited to only a few studies.…”

Complex trait–environment relationships underlie the structure of Wisconsin forest plant communities

“…Species’ performances typically show unimodal responses to environmental variables reflecting intermediate optima along the gradient (Curtis 1959; Whittaker 1967; Ter Braak & Prentice 1988; Austin 1999). Similarly, we expect unimodal performance responses to traits when functional trade-offs lead to optimum trait values within sites (Muscarella & Uriarte 2016; Rolhauser & Pucheta 2017; Rolhauser et al 2019). Environment‒performance relationships can also be bimodal because of intense negative interspecific interactions (Mueller-Dombois & Ellenberg 1974; Minchin 1987; Austin 1999), nonlinear trait‒environment relationships (Laughlin & Joshi 2015), or unmeasured environmental factors (Austin et al 1984).…”