Trait Correlations Equalize Spread Velocity Across Plant Life Histories

UC Merced UC Merced Previously Published Works

Title Trait correlations equalize spread velocity across plant life histories

Permalink https://escholarship.org/uc/item/1tn571nk

Journal GLOBAL ECOLOGY AND BIOGEOGRAPHY, 26(12)

ISSN 1466-822X

Authors Lustenhouwer, Nicky Moran, Emily V Levine, Jonathan M

Publication Date 2017-12-01

DOI 10.1111/geb.12662

License https://creativecommons.org/licenses/by-nc-nd/4.0/ 4.0

Peer reviewed

eScholarship.org Powered by the California Digital Library University of California Received: 23 June 2017 | Revised: 23 August 2017 | Accepted: 25 August 2017 DOI: 10.1111/geb.12662

RESEARCH PAPERS

Trait correlations equalize spread velocity across plant life histories

Nicky Lustenhouwer1 | Emily V. Moran2 | Jonathan M. Levine1

1Institute of Integrative Biology, ETH Zurich, Universitätstrasse 16, 8092 Zurich, Abstract Switzerland Aim: Forecasting species migration with climate change and the advance of biological invasions 2 Life & Environmental Sciences, University requires a better understanding of species’ relative migration capacity. Although theory predicts of California Merced, 5200 Lake Road, Merced, California that species combining high fecundity and dispersal with early maturation should spread the fastest, possible correlations between these traits greatly complicate predictions of species’ relative Correspondence spread velocity. We asked whether the demographic and dispersal rates controlling plant popula- Nicky Lustenhouwer, Institute of tion spread are correlated across species, and which observed association of these traits leads to Integrative Biology, ETH Zurich, Universitätstrasse 16, 8092 Zurich, the fastest spread. Switzerland. Location: Worldwide. Email: [email protected] Time period: Current. Funding information Swiss National Science Foundation, Grant/ Major taxa studied: Eighty species of herbaceous and woody plants from 35 families and 64 Award Number: 31003A_141025; UC genera. Merced Methods: We examined the relationships between age at maturity, dispersal and fecundity for 80 Editor: Greg Jordan plant species, ranging from annual herbs to trees. We incorporated these rates into a model predicting spread velocities, in order to estimate species’ spread capacity as a function of their life history.

Results: Across species, age at maturity was positively associated with both dispersal and fecundity. Given that these traits have opposing effects on spread, our models predict that species widely spaced along an age-at-maturity gradient should spread at comparable rates. This result was driven by variation between rather than within life-forms; the traits controlling spread were not correlated within annual herbs, perennial herbs or trees. The predicted spread velocities for these plant life-forms overlapped considerably, although on average, trees were predicted to spread faster than herbaceous species.

Main conclusions: Our results suggest that very different plant life histories allow for similar rates of biological invasion or native species migration under climate change. Determining where species fall within the correlated suite of traits controlling spread might provide the most effective way to predict relative spread velocities.

KEYWORDS age at maturity, demography, dispersal, fecundity, migration, population spread, range expansion

1 | INTRODUCTION Roy, & Thomas, 2011; Lenoir, Gegout, Marquet, Ruffray, & Brisse, 2008; Parmesan, 2006), yet others show no evidence of migration at The persistence of many plant populations in a warming climate all (Bertrand et al., 2011; Corlett & Westcott, 2013; Zhu, Woodall, & depends on their ability to migrate (Aitken, Yeaman, Holliday, Wang, & Clark, 2012). This variation among species in the speed and extent of Curtis-McLane, 2008; Thuiller et al., 2008). Some taxa have already range shifts is expected to create no-analogue communities, where shifted their ranges to track climate change (Chen, Hill, Ohlemuller,€ novel interactions may have large impacts on species persistence

Global Ecol Biogeogr.2017;1–10. wileyonlinelibrary.com/journal/geb VC 2017 John Wiley & Sons Ltd | 1 2 | LUSTENHOUWER ET AL.

(Alexander, Diez, & Levine, 2015; Urban, Tewksbury, & Sheldon, 2012). such as individual-based simulations (Harris, Stanford, Edwards, Travis, Variation in spread among taxa is also a prominent feature of biological & Park, 2011), analytical integrodifference equations (Caplat, Nathan, & invasions, with some exotic species spreading rapidly across the land- Buckley, 2012), or a combination of both (Travis, Harris, Park, & scape, whereas others advance more slowly (Pysek & Hulme, 2005). A Bullock, 2011), can be field parameterized for individual species and better understanding of species’ relative migration capacity would be a often produce accurate forecasts of migration rates. However, they are useful addition to efforts to predict the composition of future ecologi- usually restricted to one or a few species for which the extensive data cal communities and the success of biological invasions. needs are met (Urban et al., 2016). Species distribution models (SDMs) Theory shows that the highest spread velocities can be expected present an alternative approach and can make predictions with compa- for species combining far dispersal, high fecundity and a short time to rable methods for a variety of species. However, in SDMs future ranges maturity (Clark, 1998; Okubo & Levin, 1980; Skellam, 1951). However, are forecast based on climatic tolerances (Guisan & Thuiller, 2005), and empirical evidence suggests that these traits are correlated across spe- the predicted ability of species to fill that future range tends to be cies, which greatly complicates predictions of which species spread based on separate estimates of migration ability (e.g., Iverson, Schwartz, fastest. Trees, for example, often mature late (which slows spread), but & Prasad, 2004), rather than spatial population dynamics (but see J. disperse their seeds to greater distances than do most herbaceous life- Pagel & Schurr, 2012; Schurr et al., 2012). forms (Cain, Damman, & Muir, 1998; Svenning & Sandel, 2013), owing What is needed are modelling approaches that can be applied to greater height, well-developed dispersal structures (Venable & Levin, across a range of taxa, based on widely available data (Singer et al., 1983) and a lower seed terminal velocity (Endels et al., 2007). How- 2016), to provide a standardized measure of species’ spread capacity. ever, the vast majority of past studies have been based on categorical Such studies could forecast spread rates for a wider range of species plant life histories (life-forms such as herbs and trees) or proxies of dis- than possible with the complex demographic models requiring intensive persal ability (seed dispersal mode or terminal velocity) only, and an field parameterization (e.g., Jongejans et al., 2008), while still taking exact quantification of relationships between the traits controlling into account species-specific demographic and dispersal rates. Indeed, spread across species with different life histories is still lacking. If dis- Hemrova, Bullock, Hooftman, White, and Munzbergov€ a(2017)recently persal, fecundity and time to maturity are indeed correlated across taxa incorporated demographic data and estimates of wind dispersal into and life-forms, the key question becomes, which observed combination mathematical models of population spread to predict the invasion of these traits spreads the fastest? For example, will an annual plant velocity of 16 short-lived grassland herbs. Their finding that early matu- with early maturation, but moderate fecundity and dispersal advance ration was associated with faster spread merits exploration across the its range more rapidly (or slowly) than a later-maturing shrub with full range of plant life histories, including annuals, perennials, shrubs greater fecundity and dispersal? and trees. In principle, one can answer these questions with analyses of pub- Here, we ask: (a) whether the major life history traits determining lished spread velocities based on reconstructions of historical migra- population spread (age at maturity, dispersal and fecundity) are corre- tions, or with predicted spread velocities from models parameterized lated across species; (b) which observed association of these traits with contemporary demographic and dispersal information (Clark, leads to the fastest spread; and (c) whether these correlations cause Lewis, & Horvath, 2001; Lenoir et al., 2008; McLachlan, Clark, & different plant life-forms (annual herbs, perennial herbs, shrubs and Manos, 2005; Nathan, Horvitz, et al., 2011). However, as explained in trees) to spread at different rates. To address these questions, we the next two paragraphs, applying either approach to many species, as assessed the relationship between the three determinants of popula- required for comparative studies of spread capacity, is challenging tion spread for 80 plant species ranging widely in life-form, seed disper- because of the scarcity of relevant data. sal vector and geographical location, controlling for phylogenetic Palynological and molecular studies of post-glacial expansion typi- relatedness. We then incorporated these values into a stochastic model cally focus on temperate, wind-pollinated trees (Delcourt & Delcourt, of population spread to present a standardized estimate of spread 1987; McLachlan et al., 2005); much less is known about herbaceous velocity as a function of plant species life history. species (Cain et al., 1998) or tropical plants. In contrast, recent distribution shifts driven by contemporary climate change have been described 2 | METHODS for many species, but can be difficult to compare across studies owing to differences in how these shifts were measured (as described by 2.1 | Quantifying species’ seed dispersal, age at Lenoir & Svenning, 2015). Finally, it is not clear how much the maturity and fecundity observed variation in historical migration rates across species is determined by their innate capacity to spread versus differences in their We obtained quantitative measures of dispersal, fecundity and time to responses to other factors, such as climate forcing, habitat suitability or maturity from the literature, in metrics suitable for the parameterization landscape fragmentation (Angert et al., 2011). of population spread models. Although plant functional traits such as Modelling approaches have also been used to predict the spread height (Tackenberg, Poschlod, & Bonn, 2003) and seed size (Cappuccino, of different taxa. However, because the approach differs widely across Mackay, & Eisner, 2002) also affect spread velocity, they have only an species, the predicted spread rates are not directly comparable, as indirect impact, through their effects on these three parameters. We needed for a comparative study of spread velocity. Specifically, models therefore focused our study on dispersal, fecundity and time to maturity. LUSTENHOUWER ET AL. | 3

We used key reviews containing dispersal data (Vittoz & Engler, 2007; more accurate than negative exponential kernels for some species with Willson, 1993) and the TRY database (Kattge et al., 2011) to identify a high probability of long-distance dispersal (Clark, Silman, Kern, species for inclusion, because dispersal data are the least frequently Macklin, & HilleRisLambers, 1999), the need for more spatially exten- reported in the literature. Trait values for each species were assumed to sive data would have further restricted the number of species that be spatially and temporally invariant, although we acknowledge that could be included in this study. We therefore used a negative exponen- these values will vary to some extent with environmental conditions. tial dispersal kernel for all species, to allow for consistent predictions of spread velocity that are comparable across taxa. We used the criterion 2.1.1 | Seed dispersal of a minimal R2 of .20 to ensure that this type of kernel was a feasible Dispersal data came from studies reporting the distribution of seeds representation of the empirical data. dispersed to increasing distances from a parent plant (the seed shadow, or occasionally seedling shadow), measured in field or laboratory condi- 2.1.2 | Demography: Age at maturity and fecundity tions. Dispersal estimates for 43 of the 80 species in our analysis came For each species, we obtained measures of age at maturity (average from Willson (1993), who fitted negative exponential kernels to seed age at which an individual first produces seeds) and fecundity (annual shadows from the literature. We fitted negative exponential kernels in per individual seed production) from the literature, trait databases and R (R Core Team, 2017) for an additional 37 species, whose seed shad- directly from the authors (Supporting Information Appendix S1). In con- ows were obtained from the TRY database (Kattge et al., 2011) and trast to Hemrova et al. (2017), who incorporated complete demography other literature sources (see Supporting Information Appendix S1 for estimated from matrix population models, we focused on only two all data entries with data sources). Notably, we used seed shadows demographic traits: age at maturity and fecundity, rates which are rather than mean dispersal distances, even though the latter are more widely available in the literature. Nonetheless, our use of these data frequently reported in the literature. Given that the mean dispersal dis- requires several assumptions. Specifically, we assumed that species’ tance depends not only on the shape of the dispersal kernel, but also fecundity at maturation was equal to their eventual fecundity, and we on the number of seeds dispersed, dispersal kernels are a more suitable ignored germination failure or seedling mortality (owing to the absence metric to compare dispersal and fecundity independently across of data in most cases). Estimates from multiple sources or environ- species. Following Willson (1993), we fitted a linear regression of the form ments were averaged. For age at maturity, if only the minimum and ln(y) 52mx 1 b, where y is the number of seeds dispersed to distance maximum were reported, we used an average of the two. x from the parent plant, from the mode of the seed shadow outwards, Our final data set included 80 species from 35 families and 64 gen- dropping all zero seed arrival values before analysis. The slope parame- era (Supporting Information Appendix S1, Table S1). ter m describes the rate of decrease in dispersal with distance from the parent and was used as a measure of dispersal ability comparable 2.2 | Predicting species’ spread velocity across species. Lower values of m result in more gradual dispersal To understand which observed combination of traits leads to the fast- kernels, hence further dispersal of seeds. Fitting from the mode out- est spread, we predicted each species’ spread velocity based on their ward (which was nearly always close to the parent plant) had almost no dispersal ability (the rate parameter m), fecundity (N seeds) and age at effect on the estimate of m, relative to fitting through all points (Spear- 2 maturity ( years). We used the discrete time, stochastic model of pop- man’s q 5 .94 between the two metrics). We imposed a minimal R of g .20 for inclusion in our subsequent analyses. When multiple studies of ulation spread along a linear habitat developed by Clark et al. (2001). seed shadows were available for a single species, we used the geomet- Spread is modelled as a series of advances made by the leading individ- ric mean m value. In cases where one species had several seed dispersal ual in the population. Each advance is the distance travelled by the fur- vectors, we analysed the seed shadow only for the most effective dis- thest dispersed seed produced by the parent. Each generation of persal mechanism (i.e., the mechanism with the lowest m). spread reflects a different finite sample of the dispersal kernel, and the It should be noted that rare long-distance dispersal events were spread velocity becomes the average displacement of the leading indi- unlikely to be captured in our data sources, because the longest disper- vidual, scaled by the time to maturity (Clark et al., 2001). sal distance measured depended on the furthest distance surveyed (Ell- To estimate the spread velocity with this model structure, we strand, 2014). Moreover, experiments are usually designed to capture simulated the dispersal of N seeds produced by the leading individual, seed dispersal by the main dispersal vector, but seeds may simultane- by sampling N times from a negative exponential dispersal kernel with ously be dispersed through multiple non-standard dispersal mecha- rate parameter m, assuming an equal probability of moving forwards or nisms (Chambers & MacMahon, 1994; Higgins, Nathan, & Cain, 2003). backwards. This generates a distribution of distances the N seeds dis- For example, small seeds that are assumed to be dispersed by wind persed, the furthest forward of which describes the advance of the may also attach to the fur of animals or be transported over long dis- population in that generation. We repeated this procedure 10,000 tances via water or human transport corridors. As a consequence, the times and calculated the mean displacement of the furthest dispersed relationship between short-distance dispersal by the main dispersal seed, which when divided by the age at maturity gives the predicted vector and the probability of long-distance dispersal events may be spread velocity in metres per year (see Supporting Information poor (Higgins et al., 2003). Although fat-tailed dispersal kernels may be Appendix S2 for a worked out example). 4 | LUSTENHOUWER ET AL.

One of the advantages of the Clark et al. (2001) model is that it Finally, we grouped all species by plant life-form [annual herb incorporates the finite nature of individuals in populations. It should be (n 5 17), perennial herb (n 5 33), shrub (n 5 5) or tree (n 5 25)] and noted that, as in any simple model of spread, the velocities predicted tested whether these life-forms differed in their predicted spread may not accurately reflect the advance of these species in their real velocity (phylogenetically explicit ANOVA). For all life-form groups landscapes, because the model makes simplifying assumptions, such as except shrubs, we also investigated the correlations between age at the absence of Allee effects, interspecific competition and stochasticity maturity, dispersal and fecundity, and between each trait and the in seed production. Importantly, species are assumed to spread into spread velocity. We analysed these correlations separately, instead of continuously favourable habitat. Although we acknowledge that the incorporating life-form as a factor in our regression analyses, because realized spread velocity of species depends on the availability of suita- annuals do not vary in age at maturity, and our data set contained only ble habitat as well as species’ dispersal and life history traits (Jordan, five shrub species. We used Kendall’s s rank correlation to analyse the 2001; Pachepsky & Levine, 2011; Worth et al., 2017), species-specific within-life-form relationships between traits. The uneven distribution habitat preferences were outside the scope of the present study. of species per life-form across the phylogeny (especially trees, consist- Rather, we use the model to generate a standardized measure of ing of gymnosperm conifers and angiosperm broad-leaved species) did spread potential that compares intrinsic migration ability across diverse not allow us to use a phylogenetic regression as previously. We used taxa based on readily available trait data. non-parametric statistics rather than linear regression owing to the distribution of our data within life-form. The p-values for Kendall’s s were 2.3 | Statistical analysis approximated with a correction for tied values (Kendall package; Kendall, 1976; McLeod, 2011), and 95% confidence intervals were We first evaluated the relationship between age at maturity, dispersal computed using the NSM3 package in R (Hollander, Wolfe, & Chicken, (the rate parameter m) and fecundity across the 80 species using linear 2014; Schneider, Chicken, & Becvarik, 2015). regression. We then evaluated the relationship between each of these traits and the spread velocity, by regressing the predicted spread veloc- | ity against age at maturity, fecundity or dispersal. All variables were 3 RESULTS log10-transformed before analysis. We conducted univariate regressions rather than a multiple regression because we aimed to quantify We found that across the 80 plant species, age at maturity, which the relationship between the spread velocity and each predictor vari- slows spread, was positively associated with the two traits accelerating able, given their observed correlation with the other two. In other spread, namely dispersal and fecundity (Figure 1a,b). The association words, although the theoretical relationship between any one trait and between age at maturity and dispersal (slope 5 1.01, F1,78 5 46.7, < the predicted spread velocity was embedded in our model, the associa- p .001; Figure 1a) was far stronger than the association between age tion with the observed values for the other two traits determined what at maturity and fecundity (slope 5 0.69, F1,78 5 6.74, p 5 .01; Figure 1b) this relationship looked like across species. and was consistent across seed dispersal vectors (Supporting Informa- In all analyses, we corrected for the effect of phylogenetic non- tion Appendix S4) and across species with or without naturalized popu- independence on the relationships between variables (full methods in lations outside their native range (Supporting Information Appendix S5). Supporting Information Appendix S3), because closely related taxa Fecundity and dispersal, both of which increase spread, were positively < might share similar traits simply by virtue of their shared evolutionary associated (slope 5 0.26, F1,78 5 18.8, p .001; Figure 1c). In contrast, history. We first constructed a phylogenetic hypothesis of the evolu- there was no significant correlation between age at maturity, dispersal tionary relationships among our taxa. We assembled a backbone phy- or fecundity within plant life-forms (Table 1), indicating that the signifi- logeny with Phylomatic (Webb & Donoghue, 2005) using Phylomatic cant cross-species correlations were driven by relationships between megatree R20120829 for plants, which is based on the Angiosperm more than within life-forms. Overall, the species in our data set lie along Phylogeny Group base tree (Stevens, 2001; Webb & Donoghue, 2005). a continuum of traits relevant to spread, with annual herbs, perennial Polytomies (i.e., nodes in the tree with more than two descendent line- herbs, shrubs and trees roughly sorting along the age-at-maturity axis. ages) were resolved following the literature concerning specific families At one end of this life history continuum are species that mature early or genera (Supporting Information Appendix S3). We then used phylo- but usually have limited dispersal and fecundity, whereas at the other genetic least squares (PGLS) regression for each relationship described end are species that take a long time to mature but produce more and in the previous paragraph (Felsenstein, 1985; Revell, 2010). Following better-dispersed seeds. Revell (2010), we estimated the phylogenetic signal in the correlation As a result of these associations between traits, there was no sig- between variables (Pagel’s k; M. Pagel, 1999) and the regression coeffi- nificant relationship between age at maturity and predicted spread cients simultaneously. We assessed the statistical significance of the velocity (F1,78 5 0.45, p 5 .50; Figure 2a), meaning that species across estimated phylogenetic signal by comparing the fitted model with mod- the age-at-maturity gradient were predicted to spread at similar rates els where k was fixed at either zero (phylogenetic independence) or (median predicted spread velocity 5.28 m year21;Figure2a).Thisresult one (trait evolution following Brownian motion) using likelihood ratio is in contrast with what we would expect in the absence of between- tests. All statistical analyses were performed in R version 3.4.0 (R Core species correlations in demography and dispersal (dashed line in Team, 2017). Figure 2a; predicted spread velocity if dispersal and fecundity are set LUSTENHOUWER ET AL. | 5

(a) 0.01 spread velocity (slope 5 0.76, F1,78 5 146.5 and slope 5 0.95, F1,78 5 48.6, respectively, p < .001; Figure 2b,c). These relationships were far less 0.03 affected by the correlated nature of dispersal, fecundity and age at matu- 0.10 rity, lying closer to the relationship predicted in the absence of any associa-

m tions between these traits (Figure 2b,c). These positive relationships were 0.32 also generally significant within life-forms (Table 1). 1.00 Most relationships were phylogenetically independent (Supporting 3.20 p < 0.001 Information Table S2), although for the relationships between dispersal 10.0 = 0.39 n.s. and fecundity (Figure 1c) and between dispersal and spread velocity 1 3.2 10 32 (Figure 2b) a model incorporating phylogenetic signal provided a better Age at maturity fit than a model assuming phylogenetic independence (Supporting Information Table S2). Finally, we note the large variance in predicted (b) 108 spread velocity among species, with predictions ranging from 0.11 to 402 m year21 (Figure 2). Despite the correlations between the spread- 106 related traits shown in Figure 1, some species do combine high dispersal and early maturation and, as would be expected, these spread the 104 fastest.

Fecundity Despite the absence of an association between age at maturity and 102 p = 0.01 spread velocity across all 80 species, we found that when species were grouped by life-form, these differed in their predicted spread velocity 0 = 0.10 n.s. [F3,76 5 3.29, p 5 .025, k 5 .27 (not significantly different from zero)]. 1 3.2 10 32 Specifically, trees were predicted to spread faster than both annual and Age at maturity perennial herbs (Figure 3). The spread velocities of annual herbs, peren- (c) nial herbs and shrubs were not significantly different from each other. 0.01

0.03 4 | DISCUSSION 0.10 m 0.32 We found that the three life history traits regulating spread (dispersal, fecundity and age at maturity) are strongly associated across 80 plant 1.00 species from a wide range of taxa, life-forms and geographical regions. 3.20 p < 0.001 One result of these associations is that species varying vastly in their 10.0 = 0.72 *** maturation time are predicted to spread at similar velocities. Although 0 102 104 106 108 late-maturing species benefit from better dispersal and, to a much Fecundity lesser extent, greater annual fecundity, species that mature early com- herbaceous herbaceous pensate with their short generation time. shrub tree annual perennial Most of the patterns that we observed were driven by relation-

FIGURE 1 Relationships between the traits determining spread: ships between rather than within life-forms. Although across all taxa, (a) age at maturity and dispersal (rate parameter m,analysed age at maturity was positively correlated with dispersal (Figure 1a), we negative ensuring that higher values represent better-dispersing found no such correlation within life-form groups (Table 1). One conse- plants), (b) age at maturity and fecundity, and (c) fecundity and dis- quence of this is that later-maturing perennial herbs were predicted to persal (log scale used for all axes). Lines represent phylogenetic 10 spread more slowly than those with a shorter generation time and simi- least squares regressions (PGLS) 6 SE of the predictions. p-values refer to the statistical significance of the slope, and Pagel’s k indi- lar dispersal (Table 1), a result which matches that of Hemrovaetal. cates the estimated phylogenetic signal and whether this signal (2017) for dry-grassland herbs. They found that early-maturing herbs was significantly different from zero were predicted to spread fastest, which is expected if early maturation is not associated with a dispersal disadvantage. The five species pre- to the cross-species median). Thus, the absence of a relationship between dicted to spread the fastest in our study [the herbs Tussilago farfara age at maturity and predicted spread velocity is caused by the positive (coltsfoot), Cirsium vulgare (common thistle) and Carduus nutans (musk association between age at maturity and both dispersal and fecundity (Fig- thistle), the shrub Calluna vulgaris (heather) and the tree Melaleuca quin- ure 1b,c). We did find a negative correlation between age at maturity and quenervia (paperbark tree)] combine early maturation, high dispersal spread within the perennial herbs (Table 1), but this pattern was over- and moderate to high fecundity, which are traits favourable for spread. whelmed by the between-life-form differences in the analysis of all species. Perhaps not incidentally, all of these species have naturalized popula- Across all species, dispersal and fecundity were positively related to the tions outside their native range (Supporting Information Appendix S5). 6 | LUSTENHOUWER ET AL.

TABLE 1 Within-life-form correlations between the traits determining spread and between each of these traits and the predicted spread velocity

Herbaceous annuals Herbaceous perennials Trees 95% Confidence 95% Confidence 95% Confidence sa interval p-value s interval p-value s interval p-value

Trait correlations between dispersal and age at maturity .12 [2.10, .34] .35 .22 [2.04, .47] .16 fecundity and age at maturity 2.11 [2.37, .16] .42 2.10 [2.43, .24] .52 dispersal and fecundity .13 [2.28, .54] .48 .18 [2.05, .41] .15 0 [2.27, .27] 1

Correlations between spread velocity and age at maturity –– – 2.26 [2.51, 2.02] .04 2.22 [2.53, .10] .15 dispersal .73 [.51, .96] < .001 .61 [.40, .83] < .001 .63 [.36, .78] < .001 fecundity .41 [.06, .75] .03 .41 [.19, .63] < .001 .23 [2.10, .56] .11

Note. No age-at-maturity correlations are presented for annuals. aKendall’s s rank correlation, with 95% confidence interval for a5.05 and approximated p-value with a correction for tied values (Kendall, 1976). Bold values indicate correlations for which p < .05.

These results highlight that the relationships documented here do not than on species’ intrinsic capacity to migrate based on their dispersal result from hard ecophysiological constraints. For example, although and life history traits, as explored here (Worth et al., 2017). Possible taller plants may generally disperse better than their shorter counter- variation among species in their establishment rates in different habi- parts (Thomson, Moles, Auld, & Kingsford, 2011), some shorter plants tats therefore adds considerable uncertainty to our model-based pro- achieve effective dispersal via specialized seed traits promoting wind jections of spread. Moreover, in a climate change context, we expect dispersal. habitat suitability to change over time, further complicating the projec- Based on the lack of relationship between age at maturity and tion of realized spread. For example, taxa that spread more slowly than spread, one would predict that annual herbs, perennial herbs, shrubs their shifting climate envelope may encounter less favourable environ- and trees would all spread at comparable rates. Yet, when we grouped mental conditions in the future, which could strongly decrease their our taxa by life-form, we found that trees spread faster on average rate of spread. For our comparative study of species’ intrinsic spread than herbaceous annuals or perennials (Figure 3). The reason for this capacity, it is most important to consider how changing habitat suitabil- discrepancy is that herbaceous perennials are disproportionately repre- ity under climate change affects taxa with different life history and dis- sented among the slowest spreading, late-maturing taxa (with trees persal traits. Our expectation is that this effect would be rather making up the faster-spreading late-maturing taxa; Figure 2a). Thus, ideosynchratic and therefore add further noise to the relationships singling out trees in a life-form-based analysis reveals the faster spread between these traits and the spread velocity, making it even more diffi- of this late-maturing group. To put it another way, the dispersal and cult to use individual traits to predict spread. fecundity traits of trees better compensate for their late maturation In addition to habitat suitability, climate change may also affect dis- than do those of late-maturing herbs. persal, fecundity and age at maturity of individual species. Changes in air temperature and wind speed affect dispersal (Bullock et al., 2012; 4.1 | Key assumptions Kuparinen, Katul, Nathan, & Schurr, 2009; Nathan, Horvitz, et al.,

Our analysis is based on the premise that we need a consistent model- 2011; Nathan, Katul, et al., 2011), and rising temparatures and CO2 ling approach for projecting spread across species if these predictions concentrations could increase fecundity (Hampe, 2011; Ladeau & Clark, are then to be used in a comparative analysis. For this reason, and 2006a; Nathan, Katul, et al., 2011) and advance maturation (Ladeau & because of the limited availability of data in the literature, our spread Clark, 2006b). However, the quantitative effect of climate change on velocity predictions necessarily ignore some dispersal and demographic absolute population spread rates is expected to be moderate (Nathan, processes. Nonetheless, although the excluded factors should affect Katul, et al., 2011; Soons, Nathan, & Katul, 2004), and we have no indi- the absolute spread velocity of species, it is not clear that they should cation that these factors should differentially affect early- versus late- disproportionately affect early- or late-maturing taxa, hence the slope maturing species, hence our qualitative conclusions. of the relationship between maturation time and spread velocity shown Other assumptions could have a larger effect on some plant life- in Figure 2a. forms than others. Land-use change is expected to cause increasing For example, the inclusion of rare long-distance dispersal events habitat fragmentation in the future (Fischlin et al., 2007), which may would increase absolute spread velocities across all taxa. Non-standard affect species with animal-dispersed seeds in particular (Herrera, de Sa mechanisms of dispersal could assist the long-distance dispersal of Teixeira, Rodríguez-Perez, & Mira, 2016). Late-maturing species with seeds from both herbaceous and woody species, via mechanisms that well-dispersed seeds might be expected to overcome gaps between operate independently of seed morphology (Higgins et al., 2003). If suitable habitat more easily than their short-lived counterparts seeds are dispersed over long distances, the realized spread velocity of with poorer dispersal, meaning that their relative spread capacity in species may depend more strongly on the availability of suitable habitat fragmented landscapes may be underestimated. In contrast, our LUSTENHOUWER ET AL. | 7

) 1,000 a a ab b

(a) 1 1,000 ) 1 100 yr 100 10

10 1

Spread velocity (m yr Spread velocity 0.1 1 p = 0.50 annual perennial shrub tree

Spread velocity (m Spread velocity herb herb 0.1 = 0.34 n.s.

1 3.2 10 32 FIGURE 3 Distribution of predicted spread velocities (log10 scale) Age at maturity for annual herbs (n 5 17), perennial herbs (n 5 33), shrubs (n 5 5) and trees (n 5 25). Horizontal lines indicate the back-transformed (b) 1,000 mean logged spread velocity. Life-forms that do not share letters ) 1 spread at significantly different rates (phylogenetically explicit r 100 t tests on log10-transformed data, a5.05); trees spread significantly faster than annual herbs (t 523.04, p 5 .003) and perennial herbs (t 522.43, p 5 .017) 10

favour late-maturing taxa, with the net effect difficult to assess. 1 Despite these assumptions and limits to the data available, our esti- p < 0.001 ’

Spread velocity (m y (m Spread velocity mates of species intrinsic spread capacity produce associations 0.1 = 0.52 * between life history and spread qualitatively consistent with those 10.0 3.20 1.00 0.32 0.10 0.03 0.01 based on observations of historical migration rates, as we detail in the m next section. (c) 1,000 ) 1 r | 100 4.2 Comparison to historical migration rates As with our model-based analysis, available historical migration rates 10 for plants do not provide clear evidence for the superior migration ability of species with particular life history strategies. Distribution shifts in 1 response to recent climate change have been documented for both p < 0.001 herbaceous and woody species, especially along elevational gradients Spread velocity (m y Spread velocity 0.1 = 0.23 n.s. in mountain ecosystems. Upslope expansion rates reported across 0 2 4 6 8 10 10 10 10 studies show considerable overlap between herbaceous (2–90 m per Fecundity decade; Holzinger, Hulber,€ Camenisch, & Grabherr, 2007; Le Roux &

FIGURE 2 Predicted spread velocities along (a) an age-at-maturity McGeoch, 2008; Lenoir et al., 2008; Parolo & Rossi, 2008; Speed, gradient, (b) a dispersal gradient and (c) a fecundity gradient (log10 Austrheim, Hester, & Mysterud, 2012; Walther, Beißner, & Burga, scale). As in Figure 1, colours indicate plant life-forms, and continu- 2005) and woody plants (24–65 m per decade; Beckage et al., 2008; ous lines and statistics represent PGLS results. Dashed lines repre- Kelly & Goulden, 2008; Kullman, 2002; Lenoir et al., 2008). Likewise, sent our theoretical expectations of the relationships, predicting studies including both types of species have found equal (Felde, Kapfer, the spread velocity considering the variation in one trait only, while the other two traits are set equal for all species (to the & Grytnes, 2012; unpublished analysis of data from Telwala, Brook, median of the data set). Species along an age-at-maturity gradient Manish, & Pandit, 2013) or larger (Lenoir et al., 2008; Pucko, Beckage, did not significantly differ in their predicted spread velocity (contin- Perkins, & Keeton, 2011) shifts in altitudinal range for herbs and uous line in a), deviating strongly from the theoretical expectation grasses in comparison to shrubs and trees. Individual studies thus that age at maturity by itself has a negative effect on spread report a variety of spread velocities, which when taken together over- (dashed line in a) lap considerably between life-forms. assumptions that plants reach full fecundity at maturation and that Although differences in methodology (Lenoir & Svenning, 2015), fecundity is not reduced owing to germination failure or seedling mor- historical conditions and the spatial scale of studies (Pysek & Hulme, tality may have disproportionately elevated the spread velocity of late- 2005) may have contributed to this pattern, it is also in line with several maturing species, given that these species have a longer period of vul- features of our results. First, we found wide distributions of spread nerability between germination and maturity. velocities for each plant life-form (Figure 3), which matches the In sum, some of our data constraints and assumptions are likely to variation in observed migration rates in the literature. Second, the favour the spread of early-maturing taxa, whereas others are likely to distributions of the individual life-forms overlapped to a great extent 8 | LUSTENHOUWER ET AL.

(Figure 3), in correspondence to the overlapping ranges of migration REFERENCES rates reported for herbaceous and woody species. One difference Aitken, S. N., Yeaman, S., Holliday, J. A., Wang, T., & Curtis-McLane, S. between the approaches is that the historical migration rates represent (2008). Adaptation, migration or extirpation: Climate change out- comes for tree populations. Evolutionary Applications, 1,95–111. the realized spread of species in habitat of a certain quality over a cer- Alexander, J. M., Diez, J. M., & Levine, J. M. (2015). Novel competitors tain time period, whereas our trait-based predictions generate an intrin- shape species’ responses to climate change. Nature, 525, 515–518. sic migration capacity of species in a continuously favourable Angert, A. L., Crozier, L. G., Rissler, L. J., Gilman, S. E., Tewksbury, J. J., & landscape. Combined, these results suggest that different plant life- Chunco, A. J. (2011). Do species’ traits predict recent shifts at forms along an age-at-maturity gradient may truly have equivalent expanding range edges? Ecology Letters, 14, 677–689. migration capacities, arising from correlations among the traits control- Beckage, B., Osborne, B., Gavin, D. G., Pucko, C., Siccama, T., & Perkins, ling spread. T. (2008). A rapid upward shift of a forest ecotone during 40 years of warming in the Green Mountains of Vermont. Proceedings of the National Academy of Sciences USA, 105, 4197–4202. 4.3 | Predicting the spread velocity of species in Bertrand, R., Lenoir, J., Piedallu, C., Riofrío-Dillon, G., de Ruffray, P., the future Vidal, C., ... Gegout, J.-C. (2011). Changes in plant community composition lag behind climate warming in lowland forests. Nature, 479, With global climate change, ecologists have increasing need to forecast 517–520. the range dynamics of many species, and using categorical life-forms or Bullock, J. M., White, S. M., Prudhomme, C., Tansey, C., Perea, R., & single plant traits as proxies for the spread velocity is a potentially tan- Hooftman, D. A. P. (2012). Modelling spread of British wind- talizing option (Hemrova et al., 2017). However, based on the correla- dispersed plants under future wind speeds in a changing climate. tions between traits controlling spread, determining where species fall Journal of Ecology, 100, 104–115. within this correlated suite of traits may be a more effective way to Cain, M. L., Damman, H., & Muir, A. (1998). Seed dispersal and the Holocene migration of woodland herbs. Ecological Monographs, 68, predict relative spread velocities (Estrada, Morales-Castilla, Caplat, & 325–347. Early, 2016). We would expect, for example, relatively high spread Caplat, P., Nathan, R., & Buckley, Y. M. (2012). Seed terminal velocity, velocities for species lying above the fitted relationship between age at wind turbulence, and demography drive the spread of an invasive maturity and dispersal in Figure 1a. A similar approach, applied to many tree in an analytical model. Ecology, 93, 368–377. taxa, and in combination with species distribution models, might help Cappuccino, N., Mackay, R., & Eisner, C. (2002). Spread of the invasive to forecast the extent to which species can spread into newly available alien vine Vincetoxicum rossicum: Tradeoffs between seed dispersabil- – habitat and the composition of future no-analogue communities. Addi- ity and seed quality. The American Midland Naturalist, 148, 263 270. tionally, knowledge about trait correlations not only explains why indi- Chambers, J. C., & MacMahon, J. A. (1994). A day in the life of a seed: Movements and fates of seeds and their implications for natural and vidual traits may correlate poorly with observed spread velocities managed systems. Annual Review of Ecology and Systematics, 25, (Angert et al., 2011), but can also be used to infer missing trait informa- 263–292. tion (Santini et al., 2016). Ultimately, more precise predictions of spread Chen, I.-C., Hill, J. K., Ohlemuller,€ R., Roy, D. B., & Thomas, C. D. (2011). velocities will require species-level estimates of demography and dis- Rapid range shifts of species associated with high levels of climate – persal, an increasingly feasible prospect with the rapid growth of trait warming. Science, 333, 1024 1026. databases (Kattge et al., 2011). Clark, J. S. (1998). Why trees migrate so fast: Confronting theory with dispersal biology and the paleorecord. The American Naturalist, 152, 204–224. ACKNOWLEDGMENTS Clark, J. S., Lewis., & Horvath, L. (2001). Invasion by extremes: Popula- We thank the TRY initiative on plant traits (www.trydb.org) and H. W. tion spread with variation in dispersal and reproduction. The American – Gardner, W. Hallwachs, J. HilleRisLambers, D. H. Janzen, J. M. H. Knops, Naturalist, 157, 537 554. W.D.Koenig,W.J.PlattandD.A.Roachforprovidingdata.Wealso Clark, J. S., Silman, M., Kern, R., Macklin, E., & HilleRisLambers, J. (1999). Seed dispersal near and far: Patterns across temperate and tropical thank Rhys Ormond for his help in data collection, and the Plant Ecology forests. Ecology, 80, 1475–1494. group for helpful comments on the manuscript. N.L. and J.M.L. acknowl- Corlett, R. T., & Westcott, D. A. (2013). Will plant movements keep up edge support from the Swiss National Science Foundation grant with climate change? Trends in Ecology and Evolution, 28, 482–488. 31003A_141025, and E.V.M. from UC Merced start-up funding. Delcourt, P. A., & Delcourt, H. R. (1987). Long-term forest dynamics of the temperate zone: A case study of late-Quaternary forests in eastern North America (Vol. 63). New York, NY: Springer. DATA ACCESSIBILITY Ellstrand, N. C. (2014). Is gene flow the most important evolutionary The age at maturity and fecundity data used in this study, as well as force in plants? American Journal of Botany, 101, 737–753. all estimated m values and predicted spread velocities, are available Endels, P., Adriaens, D., Bekker, R. M., Knevel, I. C., Decocq, G., & in Supporting Information Appendix S1, with references to individual Hermy, M. (2007). Groupings of life-history traits are associated with data sources for each data entry. distribution of forest plant species in a fragmented landscape. Journal of Vegetation Science, 18, 499–508. Estrada, A., Morales-Castilla, I., Caplat, P., & Early, R. (2016). Usefulness ORCID of species traits in predicting range shifts. Trends in Ecology and Evo- Nicky Lustenhouwer http://orcid.org/0000-0002-5157-857X lution, 31, 190–203. LUSTENHOUWER ET AL. | 9

Felde, V. A., Kapfer, J., & Grytnes, J.-A. (2012). Upward shift in eleva- plants. Proceedings of the Royal Society B: Biological Sciences, 276, tional plant species ranges in Sikkilsdalen, central Norway. Ecography, 3081–3087. https://doi.org/10.1098/rspb.2009.0693 35, 922–932. Ladeau, S. L., & Clark, J. S. (2006a). Elevated CO2 and tree fecundity: Felsenstein, J. (1985). Phylogenies and the comparative method. The The role of tree size, interannual variability, and population heteroge- American Naturalist, 125,1–15. neity. Global Change Biology, 12, 822–833. Fischlin, A., Midgley, G. F., Hughs, L., Price, J., Leemans, R., Gopal, B., ... Ladeau, S. L., & Clark, J. S. (2006b). Pollen production by Pinus taeda

Velichko, A. A. (2007). Ecosystems, their properties, goods and serv- growing in elevated atmospheric CO2. Functional Ecology, 20, ices. In M. L. Parry, O. F. Canziani, J. P. Palutikof, P. J. van der Lin- 541–547. den, & C. E. Hanson (Eds.), Climate change 2007: Impacts, adaptation Le Roux, P. C., & McGeoch, M. A. (2008). Rapid range expansion and and vulnerability. Contribution of working group II to the fourth assess- community reorganization in response to warming. Global Change ment report of the intergovernmental panel on climate change (pp. Biology, 14, 2950–2962. 212–272). Cambridge, U.K.: Cambridge University Press. Lenoir, J., Gegout, J. C., Marquet, P. A., Ruffray, P. D., & Brisse, H. Guisan, A., & Thuiller, W. (2005). Predicting species distribution: Offering (2008). A significant upward shift in plant species optimum elevation more than simple habitat models. Ecology Letters, 8, 993–1009. during the 20th century. Science, 320, 1768–1771. Hampe, A. (2011). Plants on the move: The role of seed dispersal and Lenoir, J., & Svenning, J.-C. (2015). Climate-related range shifts – A initial population establishment for climate-driven range expansions. global multidimensional synthesis and new research directions. Ecog- Acta Oecologica, 37, 666–673. raphy, 38,15–28. Harris, C. M., Stanford, H. L., Edwards, C., Travis, J. M. J., & Park, K. J. McLachlan, J. S., Clark, J. S., & Manos, P. S. (2005). Molecular indicators (2011). Integrating demographic data and a mechanistic dispersal of tree migration capacity under rapid climate change. Ecology, 86, model to predict invasion spread of Rhododendron ponticum in differ- 2088–2098. ent habitats. Ecological Informatics, 6, 187–195. McLeod, A. I. (2011). Kendall: Kendall rank correlation and Mann-Kendall Hemrova, L., Bullock, J. M., Hooftman, D. A. P., White, S. M., & trend test. R package version 2.2. Retrieved from http://CRAN.R-pro- Munzbergov€ a, Z. (2017). Drivers of plant species’ potential to spread: ject.org/package5Kendall The importance of demography versus seed dispersal. Oikos, 126, 1493–1500. Nathan, R., Horvitz, N., He, Y., Kuparinen, A., Schurr, F. M., & Katul, G. G. (2011). Spread of North American wind-dispersed trees in future Herrera, J. M., de Sa Teixeira, I., Rodríguez-Perez, J., & Mira, A. (2016). environments. Ecology Letters, 14, 211–219. Landscape structure shapes carnivore-mediated seed dispersal kernels. Landscape Ecology, 31, 731–743. Nathan, R., Katul, G. G., Bohrer, G., Kuparinen, A., Soons, M. B., Thomp- son, S. E., ... Horn, H. S. (2011). Mechanistic models of seed disper- Higgins, S. I., Nathan, R., & Cain, M. L. (2003). Are long-distance dispersal sal by wind. Theoretical Ecology, 4, 113–132. events in plants usually caused by nonstandard means of dispersal? Ecology, 84, 1945–1956. Okubo, A., & Levin, S. A. (1980). Diffusion and ecological problems: Mathe- matical models. Berlin, Germany: Springer. Hollander, M., Wolfe, D. A., & Chicken, E. (2014). The independence problem. In Nonparametric statistical methods (3rd ed., pp. 393–450). Pachepsky, E., & Levine, J. M. (2011). Density dependence slows Hoboken, NJ: Wiley. invaderspreadinfragmentedlandscapes.The American Naturalist, 177,18–28. Holzinger, B., Hulber,€ K., Camenisch, M., & Grabherr, G. (2007). Changes in plant species richness over the last century in the eastern Swiss Pagel, J., & Schurr, F. M. (2012). Forecasting species ranges by statistical Alps: Elevational gradient, bedrock effects and migration rates. Plant estimation of ecological niches and spatial population dynamics. Ecology, 195, 179–196. Global Ecology and Biogeography, 21, 293–304. Iverson, L. R., Schwartz, M. W., & Prasad, A. M. (2004). Potential coloniza- Pagel, M. (1999). Inferring the historical patterns of biological evolution. tion of newly available tree-species habitat under climate change: An Nature, 401, 877–884. analysis for five eastern US species. Landscape Ecology, 19, 787–799. Parmesan, C. (2006). Ecological and evolutionary responses to recent cli- Jongejans, E., Shea, K., Skarpaas, O., Kelly, D., Sheppard, A. W., & Wood- mate change. Annual Review of Ecology, Evolution, and Systematics, 37, burn, T. L. (2008). Dispersal and demography contributions to popula- 637–669. tion spread of Carduus nutans in its native and invaded ranges. Parolo, G., & Rossi, G. (2008). Upward migration of vascular plants fol- Journal of Ecology, 96, 687–697. lowing a climate warming trend in the Alps. Basic and Applied Ecology, Jordan, G. J. (2001). An investigation of long-distance dispersal based on 9, 100–107. species native to both Tasmania and New Zealand. Australian Journal Pucko, C., Beckage, B., Perkins, T., & Keeton, W. S. (2011). Species shifts of Botany, 49, 333–340. in response to climate change: Individual or shared responses? The Kattge, J., Díaz, S., Lavorel, S., Prentice, I. C., Leadley, P., Bonisch,€ G., ... Journal of the Torrey Botanical Society, 138, 156–176. Wirth, C. (2011). TRY – A global database of plant traits. Global Pysek, P., & Hulme, P. E. (2005). Spatio-temporal dynamics of plant inva- Change Biology, 17, 2905–2935. sions: Linking pattern to process. Ecoscience , 12, 302–315. Kelly, A. E., & Goulden, M. L. (2008). Rapid shifts in plant distribution R Core Team. (2017). R: A language and environment for statistical com- with recent climate change. Proceedings of the National Academy of puting. Vienna, Austria: R Foundation for Statistical Computing. Sciences USA, 105, 11823–11826. Retrieved from http://www.R-project.org/ Kendall, M. G. (1976). Rank correlation methods (4th ed.). London, U.K.: Revell, L. J. (2010). Phylogenetic signal and linear regression on species Griffin. data. Methods in Ecology and Evolution, 1, 319–329. Kullman, L. (2002). Rapid recent range-margin rise of tree and shrub spe- Santini, L., Cornulier, T., Bullock, J. M., Palmer, S. C. F., White, S. M., – cies in the Swedish Scandes. Journal of Ecology, 90,68 77. Hodgson, J. A., ... Travis, J. M. J. (2016). A trait-based approach for Kuparinen, A., Katul, G., Nathan, R., & Schurr, F. M. (2009). Increases in predicting species responses to environmental change from sparse air temperature can promote wind-driven dispersal and spread of data: How well might terrestrial mammals track climate change? 10 | LUSTENHOUWER ET AL.

Global Change Biology, 22, 2415–2424. https://doi.org/10.1111/gcb. Venable, D. L., & Levin, D. A. (1983). Morphological dispersal structures 13271 in relation to growth habit in the Compositae. Plant Systematics and – Schneider, G., Chicken, E., & Becvarik, R. (2015). NSM3: Functions and Evolution, 143,1 16. datasets to accompany Hollander, Wolfe, and Chicken – Nonpara- Vittoz, P., & Engler, R. (2007). Seed dispersal distances: A typology based on metric statistical methods, third edition. R package version 1.3. dispersal modes and plant traits. Botanica Helvetica, 117, 109–124. Retrieved from http://CRAN.R-project.org/package5NSM3 Walther, G.-R., Beißner, S., & Burga, C. A. (2005). Trends in the upward Schurr, F. M., Pagel, J., Cabral, J. S., Groeneveld, J., Bykova, O., O’Hara, shift of alpine plants. Journal of Vegetation Science, 16, 541–548. ... ’ R. B., Zimmermann, N. E. (2012). How to understand species Webb, C. O., & Donoghue, M. J. (2005). Phylomatic: Tree assembly for niches and range dynamics: A demographic research agenda for bio- applied phylogenetics. Molecular Ecology Notes, 5, 181–183. geography. Journal of Biogeography, 39, 2146–2162. Willson, M. F. (1993). Dispersal mode, seed shadows, and colonization Singer, A., Johst, K., Banitz, T., Fowler, M. S., Groeneveld, J., Gutierrez, patterns. Vegetatio, 107–108, 261–280. A. G., ... Travis, J. M. J. (2016). Community dynamics under environ- Worth, J. R. P., Holland, B. R., Beeton, N. J., Schonfeld,€ B., Rossetto, M., mental change: How can next generation mechanistic models Vaillancourt, R. E., & Jordan, G. J. (2017). Habitat type and dispersal improve projections of species distributions? Ecological Modelling, mode underlie the capacity for plant migration across an intermittent 326,63–74. seaway. Annals of Botany. https://doi.org/10.1093/aob/mcx086 Skellam, J. G. (1951). Random dispersal in theoretical populations. Biome- Zhu, K., Woodall, C. W., & Clark, J. S. (2012). Failure to migrate: Lack of trika, 38, 196–218. tree range expansion in response to climate change. Global Change Soons, M. B., Nathan, R., & Katul, G. G. (2004). Human effects on long- Biology, 18, 1042–1052. distance wind dispersal and colonization by grassland plants. Ecology, 85, 3069–3079. Speed, J. D. M., Austrheim, G., Hester, A. J., & Mysterud, A. (2012). Ele- vational advance of alpine plant communities is buffered by herbi- BIOSKETCHES – vory. Journal of Vegetation Science, 23, 617 625. NICKY LUSTENHOUWER recently completed her PhD in the Levine group at Stevens, P. F. (2001). Angiosperm phylogeny Website. Version 12, July ETH Zurich. Her dissertation work explores the effects of life history 2012 [and more or less continuously updated since]. Retrieved from and rapid evolution on spreading plant populations, in order to improve http://www.mobot.org/MOBOT/research/APweb/ forecasts of biological invasions and range expansions induced by cli- Svenning, J.-C., & Sandel, B. (2013). Disequilibrium vegetation dynamics under future climate change. American Journal of Botany, 100,1266–1286. mate change.

Tackenberg, O., Poschlod, P., & Bonn, S. (2003). Assessment of wind dis- EMILY V. MORAN is Assistant Professor at the University of California, – persal potential in plant species. Ecological Monographs, 73, 191 205. Merced, where she studies dispersal, local adaptation and the interplay Telwala, Y., Brook, B. W., Manish, K., & Pandit, M. K. (2013). Climate- between ecological and evolutionary responses to environmental induced elevational range shifts and increase in plant species richness change in plants. in a Himalayan biodiversity epicentre. PLoS One, 8, e57103.

Thomson, F. J., Moles, A. T., Auld, T. D., & Kingsford, R. T. (2011). Seed JONATHAN M. LEVINE is Professor of Plant Ecology at ETH Zurich. He dispersal distance is more strongly correlated with plant height than combines theoretical and empirical approaches to explore the mainte- with seed mass. Journal of Ecology, 99, 1299–1307. nance of species diversity, controls over plant invasions, and plant com- Thuiller, W., Albert, C., Araujo, M. B., Berry, P. M., Cabeza, M., Guisan, munity responses to climate change. A., ... Zimmermann, N. E. (2008). Predicting global change impacts on plant species’ distributions: Future challenges. Perspectives in Plant Ecology Evolution and Systematics, 9, 137–152. SUPPORTING INFORMATION Travis, J. M. J., Harris, C. M., Park, K. J., & Bullock, J. M. (2011). Improv- Additional Supporting Information may be found online in the sup- ing prediction and management of range expansions by combining porting information tab for this article. analytical and individual-based modelling approaches. Methods in Ecology and Evolution, 2, 477–488. Urban, M. C., Bocedi, G., Hendry, A. P., Mihoub, J.-B., Pe’er, G., Singer, How to cite this article: Lustenhouwer N, Moran EV, Levine A., ... Travis, J. M. J. (2016). Improving the forecast for biodiversity under climate change. Science, 353, aad8466. JM. Trait correlations equalize spread velocity across plant life – Urban, M. C., Tewksbury, J. J., & Sheldon, K. S. (2012). On a collision histories. Global Ecol Biogeogr. 2017;00:1 10. https://doi.org/ course: Competition and dispersal differences create no-analogue 10.1111/geb.12662 communities and cause extinctions during climate change. Proceedings of the Royal Society B: Biological Sciences, 279, 2072–2080.