High dispersal ability is related to fast life‐history strategies

Seed dispersal is an essential, yet often overlooked process in plant ecology and evolution, affecting adaptation capacity, population persistence and invasiveness. A species’ ability to disperse is expected to covary with other life‐history traits to form dispersal syndromes. Dispersal might be linked to the rate of life history, fecundity or generation time, depending on the relative selection pressures of bet‐hedging, kin competition or maintaining gene flow. However, the linkage between dispersal and plant life‐history strategies remains unknown because it is difficult to observe, quantify and manipulate the influence of dispersal over large spatiotemporal scales. We integrate datasets describing plant vital rates, dispersal and functional traits to incorporate dispersal explicitly into the rich spectra of plant life‐history strategies. For 141 plant species, we estimated dispersal ability by predicting maximum dispersal distances using allometric relationships based on growth form, dispersal mode, terminal velocity and seed mass. We derived life‐history traits from matrix population models parameterized with field data from the COMPADRE Plant Matrix Database. We analysed the covariation in dispersal ability and life‐history traits using multivariate techniques. We found that three main axes of variation described plant dispersal syndromes: the fast‐slow life‐history continuum, the dispersal strategy axis and the reproductive strategy axis. On the dispersal strategy axis, species’ dispersal abilities were positively correlated with aspects of fast life histories. Species with a high net reproductive rate, a long window of reproduction, low likelihood of escaping senescence and low shrinkage tendencies disperse their seeds further. The overall phylogenetic signal in our multidimensional analyses was low (Pagel's λ < 0.24), implying a high degree of taxonomic generality in our findings. Synthesis. Dispersal has been largely neglected in comparative demographic studies, despite its pivotal importance for populations. Our explicit incorporation of dispersal in a comparative life‐history framework provides key insights to bridge the gap between dispersal ecology and life‐history traits. Species with fast life‐history strategies disperse their seeds further than slow‐living plants, suggesting that longer dispersal distances may allow these species to take advantage of habitats varying unpredictably in space and time as a bet‐hedging strategy.


| INTRODUC TI ON
The plant kingdom, with over 350,000 extant species (Paton et al., 2008 ), has evolved a myriad of strategies to overcome the implications of one of its main features: sessility. By far, the most striking strategy to overcome this limitation is the dispersal of propagules such as seeds (Howe & Smallwood, 1982 ;Janzen, 1970 ;Levin, Muller-Landau, Nathan, & Chave, 2003 ). Seed dispersal, the movement of seeds away from the parent, is key in ecological and evolutionary processes (Clobert, Baguette, Benton, & Bullock, 2012 ).
Plants exhibit a variety of different methods to disperse their seeds-ingestion by animals, wind and ballistic are a few examples; these are inferred from fruit and seed morphology and are referred to as dispersal modes (Howe & Smallwood, 1982 ). Along with dispersal, plants have evolved a myriad of strategies to survive, grow and reproduce in a variety of habitats (Salguero-Gómez et al., 2016 ;Silvertown, Franco, & Harper, 1997 ). Only by understanding how dispersal covaries with life-history traits will researchers gain a more complete understanding of plant life-history strategies and the ability of plant species to respond to environmental change (Ronce & Clobert, 2012 ;Travis et al., 2013 ;Uemura & Hausman, 2013 ). In general, however, seed dispersal has traditionally been excluded from assessments of plant life-history evolution .
For plants, seed dispersal can influence fitness by determining the seedscape, that is, the abiotic and biotic environment that affects all later stages of recruitment, from seedling establishment to future reproduction (Beckman & Rogers, 2013 ). Where a seed is deposited determines the degree of competition (Loiselle, 1990 ;Spiegel & Nathan, 2012 ), the presence of natural enemies that consume plants (Connell, 1971 ;Janzen, 1970 ) and the suitability of the environment for survival and growth (Howe & Miriti, 2004 ;Schupp, Jordano, & Maria Gomez, 2010 ). The spatial template created by dispersal influences the persistence of populations and metapopulations (Jordano, 2017 ), and long-distance dispersal drives the spread of populations into new areas (Kot, Lewis, & van den Driessche, 1996 ;Neubert & Caswell, 2000 ), which is important for tracking changing climates Loarie et al., 2009 ) and species invasions (Buckley et al., 2005 ;Skarpaas & Shea, 2007 ). Dispersal is also central to the genetic diversity of populations (Hamrick, Murawski, & Nason, 1993 ) and their ability to adapt to new conditions (Kremer et al., 2012 ). Consequently, for sessile organisms such as plants, their ability to disperse and adapt to new environments will influence a species' persistence and migration into new areas under global change.
The repertoire of life-history traits in the plant kingdom is truly vast. Differential investments in maintenance result in mean life expectancies that range between weeks (e.g. Sanguinaria , Arabidopsis ) and thousands of years (e.g. Pinus longaeva , Lomantia tasmanica ; multiple phenotypic traits, including life-history or behavioural traits, as dispersal syndromes, and reserve dispersal mode for the method of dispersal inferred from fruit or seed morphology. While examining the covariation of dispersal with life-history traits, or lack thereof, cannot distinguish the underlying mechanisms or its ultimate causes, evaluating the presence or absence of dispersal syndromes in plants can help elucidate the joint evolution of traits among species and the demographic consequences of dispersal (Ronce & Clobert, 2012 ). The relationship between dispersal and plant life-history strategies has remained unclear to date due to the fact that quantifying seed dispersal empirically is challenging, and hence, data on dispersal tends to be limited. However, the availability of large volumes of open-access data on demography (Salguero-Gómez et al., 2015 ), dispersal (Bullock et al., 2017 ;Tamme et al., 2014 ) and functional traits (Kattge et al., 2011 ) is increasing.
By synthesizing available data on life-history traits, as derived from empirical stage-based demographic models, with recent approaches to predict dispersal ability from plant traits, we incorporate dispersal ability into analyses of life-history strategies and examine whether dispersal ability covaries with life-history traits, and if so, how. If dispersal is independent of life-history traits, we predict an independent axis of variation describing dispersal will be added to the existing axes describing the fast-slow continuum and the mode of reproduction. However, if dispersal ability covaries with life-history traits, then we predict that the axes of variation describing life-history strategies will shift with the inclusion of dispersal. We evaluate the following hypotheses regarding how dispersal will covary with life-history traits: 1 . If dispersal evolved as a bet-hedging strategy to take advantage of habitats that vary unpredictably in space and time, we predict dispersal distance to correlate positively with life-history traits indicative of fast life-history strategies (e.g. short generation times, high investments on reproduction, high individual growth rates; Baker & Stebbins, 1965 ;McPeek & Holt, 1992 ;Roff, 1975 ;Snyder, 2011 ).

.
Dispersal distance could also be positively correlated with generation time; species with shorter generation times can maintain gene flow through time (e.g. they have more opportunities to exchange genetic information per unit time), and species with longer generation times are expected to maintain gene flow through space (Stevens, Trochet, Van Dyck, Clobert, & Baguette, 2012 ). In addition, organisms can evolve longer life spans when their propagules disperse farther, escaping kin competition for space (Dytham & Travis, 2006 ). Growth form and plant size can constrain the variation of plant life histories, and, in combination with dispersal modes, could constrain dispersal syndromes. Growth form describes potential constraints due to anatomy (Salguero-Gómez et al., 2016 ), while dispersal mode, growth form and plant height explain dispersal distances (Tamme et al., 2014 ;Thomson, Letten, Tamme, Edwards, & Moles, 2018 ). Salguero-Gómez et al. ( 2016 ) found that growth form and matrix dimension-associated with plant size and life cycle complexity-explained the location of species on the axes of variation describing the fast-slow continuum and the mode of reproduction. We examine whether a plant ' s mode of dispersal, growth form and matrix dimension help explain how species are distributed along axes of variation describing dispersal and life-history traits. If dispersal varies independently of life-history traits, we expect dispersal mode to explain a species' location on the axis of variation that captures dispersal ability (Tamme et al., 2014 ), and growth form to explain species' locations on the axes capturing the fast-slow continuum and reproductive strategies (Salguero-Gómez et al., 2016 ), while matrix dimension would be associated with both dispersal and life-history traits.

| MATERIAL S AND ME THODS
To determine how dispersal covaries with life-history traits, we brought together data on demography, dispersal and functional traits. Demographic data were obtained from the COMPADRE Plant Matrix Database (Salguero-Gómez et al., 2015 ). We defined dispersal ability in terms of maximum dispersal distance and obtained measured maximum dispersal distances from Tamme et al. ( 2014 ).
The number of species that overlapped between both datasets was 47. To include more species and improve our ability to ascertain general patterns, we predicted maximum dispersal distances for species using information on functional traits, dispersal mode and growth form based on published relationships (Tamme et al., 2014 ); this resulted in a total of 141 species for which we were able to predict dispersal distances and derive life-history traits.
Species names across datasets were standardized using The Taxonomic Name Resolution Service using all available data sources (Boyle et al., 2013 ; The Taxonomic Name Resolution Service), including Tropicos (Missouri Botanical Garden ), USDA (USDA, NRCS ), Global Composite Checklist (Flann, 2009), The Plant List ( 2013 and the International Legume Database and Information Service . Angiosperm Phylogeny Group III was used for standardization of family and order names ( http://www.mobot.org/MOBOT/research/ APweb/ ) . Species with unresolved taxonomic names that otherwise matched across datasets were retained in the analysis. We obtained a phylogeny by trimming down the tree made available by Zanne et al. ( 2014 ) to the number of species used in each analysis using the package ape (Paradis, Claude, & Strimmer, 2004 ) in R (R Core Team, 2017 ).

| Dispersal ability
Dispersal ability in evolutionary models is usually expressed as the propensity to disperse or a summary of dispersal distances (e.g. statistical moments of distribution, non-local dispersal; Duputié & Massol, 2013 ;Saastamoinen et al., 2018 ). Maximum dispersal distance is a useful metric as it represents long-distance dispersal ability, directly relating to the ability to colonize new areas, and due to the high correlation with mean dispersal distances ( R 2 = .85-. 90 Tamme et al., 2014 ;Thomson, Moles, Auld, & Kingsford, 2011 ), it also represents local dispersal ability relevant to within population processes.
We generated maximum dispersal distances using the dispeRsal function (version 0.2) developed by Tamme et al. ( 2014 ). The function predicts maximum dispersal distances using a linear mixed-effects model that can include information on dispersal mode (i.e. no special mechanism, ballistic, ant, wind, animal), growth form (i.e. herb, shrub, tree), seed mass or terminal velocity (i.e. velocity of the diaspore falling in still air, Greene & Johnson, 1992 ) as fixed effects, as well as taxonomic family or order as random effects. We obtained trait data from open-access sources as described in the Supplementary Methods in the Supporting Information . A list of data sources from the TRY Plant Trait Database used in this study are provided in the Data sources section. We log 10 -transformed continuous trait values to meet parametric assumptions, and calculated the mean for each species if more than one value was reported.
To predict maximum dispersal distances, we chose models based on available information for each species that resulted in the highest predictive power as determined by the R 2 reported in Tamme et al.
Life-history traits were derived from MPMs (Caswell, 2001 ). where sexual reproduction had been explicitly quantified as we were interested in movement capacity via seed dispersal by seed-bearing plants. Through various demographic methods described elsewhere (Caswell, 2001 ;Cochran & Ellner, 1992 ), multiple summary statistics of vital rates and resulting life-history traits that describe the strategies of different species can be calculated from these MPMs.
Nine life-history traits were calculated sensu Salguero-Gómez et al.

| Statistical analyses
We used principal component analyses (PCA) to examine how predicted maximum dispersal distance ( D ) relates with plant life-history traits as described in Table 1 . This allowed us to examine covariation of maximum dispersal distances with life-history traits associated with fast life-history strategies (e.g. short generation times, high reproduction and high progressive growth; Hypothesis 1), high fecundity (high net reproductive rate, mean sexual reproduction; Hypothesis 2) and long generation times (Hypothesis 3). We used parallel analysis to simulate 95th percentile eigenvalues from 100 simulated analyses and determine the number of retained principal components (Franklin, Gibson, Robertson, Pohlmann, & Fralish, 1995 ;Horn, 1965 ) using the function f a.parallel in the psych package (Revelle, 2017 ) in R (R Core Team, 2017 ).
To control for non-independence among lineages in our study, we conducted a phylogenetically informed PCA using the package phytools (Revell, 2012 ) in R (R Core Team, 2017 ). This approach differs from a standard PCA in that it explicitly incorporates the phylogeny into the backbone of the PCA, correcting for the lack of independence and estimating the phylogenetic signal in the examined relationships at the same time. We estimated Pagel ' s λ, a scaling parameter that quantifies the phylogenetic correlation among species, ranging between 0 (no role of phylogeny in determining trait covariation) and 1 (trait covariation fully explained by phylogenic relationships, assuming Brownian motion; Freckleton, Harvey, & Pagel, 2002) .
We first conducted a PCA on life-history traits for the 141 species, not including predicted maximum dispersal distances. Subsequently, we conducted the PCA including life-history traits and predicted maximum dispersal distances to explicitly incorporate dispersal ability into the life-history strategies. We compared the PCA results with and without dispersal distances to examine whether and how variation in axes of life-history strategies shift when including dispersal ability. We also conducted the PCA for the 47 species that had measured maximum dispersal distances in Tamme et al. ( 2014 ). For each analysis (Tables S3 -S5 ), the phylogenetic signal as measured by Pagel ' s λ was low (<0.24), suggesting a weak effect of the ancestral relationships among the examined lineages onto the configuration of life-history traits and dispersal (Freckleton, 2012 ). Hence, here we report the results for the non-phylogenetically informed PCA.
For each PCA, we used the imputePCA function of the missMDA . We controlled for matrix dimension before including additional terms as it is a property of these demographic models known to bias demographic outputs (Enright, Franco, & Silvertown, 1995 ;Salguero-Gómez & Plotkin, 2010 ); the order of dispersal mode or growth form did not influence results and are similar to results using Type 2 SS (data not shown). Because of the small sample sizes across most combinations of growth forms and dispersal modes (Table S2 ), we only include herbs and trees that are dispersed by either wind or animals in this analysis (59 animal-dispersed herbs, 24 animal-dispersed trees, 19 wind-dispersed herbs and 15 wind-dispersed trees).
To ensure our results were not affected by statistical dependence among the principal components, we conducted a permutational ANOVA on the first three principal components using Euclidian distance and 9,999 permutations with the function adonis in the package vegan (Oksanen et al., 2017 ) in R (R Core Team, 2017 ). As the results for the permutational ANOVA (Table S9 ) were the same as the results for the ANOVA, we report the results obtained from the ANOVA here. All analyses were conducted in R (R Core Team, 2017 ).

| Life-history strategies
For life-history traits of the 141 species, the first three principal components were retained ( Figure S1 ) and explain 65% of the variation (Table S6 ). The first axis of variation, principal component one

| Do plants exhibit dispersal syndromes?
After incorporating dispersal ability into the PCA for the 141 species, the first three principal components were retained ( Figure S3 ) and capture 61% of the variation in life-history strategies (Table 1 ,   Table S7 ). The first principal component is qualitatively similar to the above analyses that do not include maximum dispersal distances, and it depicts the fast-slow continuum (Figure 1 )

| Do dispersal mode and growth form explain the distribution of species along dispersal syndromes?
Both

| DISCUSSION
Our study shows that seed dispersal ability-defined as maximum dispersal distance-has not evolved independently of other life-history traits across plant species. Dispersal seems indeed to be an integral part of a complex suite of traits; dispersal and life-history traits covary to form dispersal syndromes (Ronce & Clobert, 2012 ), which have also been recognized among animals (Stevens et al., 2014 ). In our analyses with 141 plant species, dispersal syndromes comprised three main axes of variation related to the fast-slow con-

| Life-history strategies
Our decomposition of the main drivers of life-history strategies in plants is qualitatively similar to previous examinations (Salguero-Gómez et al., 2016 ). However, a few variables have shifted their importance on each axis of life-history variation. This is most likely because the 141 species included in our analyses do not capture the range of growth forms included in the previous study of 418 species (Table S2 ). We excluded clonal species and species that did not have sexual reproduction quantified as we wanted to focus on movement ability in terms of seed dispersal by seed-bearing plants. Plant species with a clonal strategy may differ in their life-history strategies compared to non-clonal species (Kroon & Groenendael, 1997 ). For example, clonal species may have alternative strategies to escape herbivory associated with tolerance (Pellissier et al., 2016 ), different patterns of senescence due to the accumulation of mutations (Ally, Ritland, & Otto, 2010 ;Salguero-Gómez, 2017 ) and trade-offs with investment in sexual reproduction, influencing fecundity (Barrett, 2015 ).

| The dispersal syndromes of plants
We found evidence for dispersal syndromes across plant species as the axes of variation of life-history strategies based on life-history traits shifted with the inclusion of maximum dispersal distances.
Specifically, we found that high dispersal ability was related to fast life-history strategies. Previous studies in plants have provided evidence for dispersal syndromes based on phenotypic traits, such as plant height, seed mass and dispersal structures (Tamme et al., 2014 ;Thomson et al., 2011Thomson et al., , 2018. Across animal species, Stevens et al. ( 2014 ) found that dispersal ability was consistently associated with high fecundity and survival, and, in aerial dispersers, with early maturation, but the strength, direction and functional form of these relationships varied within taxonomic orders. Across animal and plant groups, species seem to have converged on similar dispersal syndromes that relate high dispersal ability with aspects of fast lifehistory strategies, including high reproductive rates for all organisms studied, early maturation for aerial-dispersing terrestrial animals and a long window of reproduction for terrestrial plants. This could have arisen from two potential mechanisms: (1) a suite of traits evolving as a bet-hedging strategy in response to unpredictable habitats (Prediction 1) or (2) dispersal evolving in response to the negative impacts of high fecundity on fitness via increased kin competition or mortality due to specialized natural enemies (Prediction 2). In each case, dispersal has evolved in tandem with life-history traits.
For this analysis, we predicted dispersal distances for 141 species using dispersal mode and growth form, including functional traits and taxonomic names when possible, to increase the number of species included in the PCA. This approach seemed to capture the covariation in dispersal and life-history traits as the results were qualitatively similar for the 47 species for which there were measured dispersal distances available in Tamme et al. ( 2014 ). In previous analyses, major habitat was a significant predictor of the T A B L E 2 Results of ANOVA including matrix dimension, dispersal mode and growth form as predictors for the first three principal components including life-history traits and dispersal mode for 117 species position of plant species on the fast-slow axis of life-history strategies (Salguero-Gómez et al., 2016 ). Dispersal ability of a plant species may also vary by habitat as the mode of dispersal can depend on precipitation, temperature and altitude (Almeida-Neto, Campassi, Galetti, Jordano, & Oliveira-Filho, 2008 ;Chen, Cornwell, Zhang, & Moles, 2017 ). As data for dispersal and demography continue to grow, future studies can examine how these relationships and dispersal syndromes vary by major habitat.
Here, we have measured but one component of the dispersal process-the maximum dispersal distance of seeds. However, dispersal is made up of several phases, including pre-departure, departure, transfer and settlement. The maximum dispersal distance of seeds is most related to the transfer phase. Each of these phases incur some costs to dispersing Clobert, Le Galliard, Cote, Meylan, & Massot, 2009 ), and selection will act to optimize fitness by minimizing costs associated across these multiple phases of dispersal Travis et al., 2012 ). Both the pre-departure and departure phases include costs associated with the investment in dispersal structures, from no specialized structures to plumes or wings for dispersal by wind to fleshy fruits for dispersal by animals. Each of these may have different energy requirements and developmental times, which may covary with other life-history traits, such as traits related to reproduction and growth.
During these phases, plants may incur costs from predators and pathogens that reduce their ability to disperse seeds (Tewksbury & Nabhan, 2001 ;Tiansawat, Beckman, & Dalling, 2017 ). A plant ' s ability to deter reductions in dispersal due to predation would be related to its ability to develop and disperse fast enough to escape predation (related to growth strategies), satiate predators (related to number of seeds produced) or otherwise defend themselves. After the settlement phase, seedlings will have to grapple with the local environment, competition with their neighbours, and mortality due to natural enemies which may be related to growth strategies (e.g. tolerance vs. defence) and turnover.
This study suggests that dispersal syndromes exist across plant species, but more work needs to be done in terms of individual variation where dispersal is predicted to be an independent axis . Saastamoinen et al. ( 2018 ) showed that dispersal is a complex process arising from several interacting traits and a complex genetic architecture; they found that although some genes influence certain aspects of dispersal with moderate to large effect, dispersal traits are typically polygenic. Studies on the genetic correlations of dispersal tend to be scarce, and the topic requires further study. In contrast, within species correlations among dispersal traits as well as between dispersal traits and other traits under selection are more common in animals (Saastamoinen et al., 2018 ), but less well studied in plants. Future research on the drivers of variance in dispersal distances within species will give additional insight into the fitness benefits of dispersal and potential trade-offs or synergies with life-history traits.
Finally, selection for longer dispersal distances will influence the ability of plant species to invade new habitats (Hastings et al., 2005 ), track changing climates (Travis et al., 2013 ) and persist in fragmented landscapes (Williams, Kendall, & Levine, 2016 (Thomson et al., 2018 ). In addition, taller trees may have a higher likelihood of dispersing further distances because they produce more seeds (Moles, Falster, Leishman, & Westoby, 2004 ). We found wind-dispersed trees had higher mean scores on the dispersal strategy axis (PC2), suggesting higher dispersal ability, compared to animal-dispersed trees. These patterns could partly be due to wind-dispersed species being taller than animal-dispersed species as plant height increases across dispersal modes (unassisted < ant < vertebrate < wind; Thomson et al., 2018 ). For wind-dispersed species, plant height per se can increase dispersal distances due to canopy wind conditions (Augspurger, Franson, Cushman, & Muller-Landau, 2016 ;Soons, Heil, Nathan, & Katul, 2004

| CONCLUSIONS
Here