Polyploidy can cause variation in plant functional traits and thereby generate individuals that can adapt to fluctuating environments and exploit new environments. However, few empirical studies have tested for an association between ploidy level and climatic tolerance of invasive cytotypes relative to conspecific native-range cytotypes. Here, we used an invasive plant Solidago canadensis to test whether invasive populations had a higher proportion of polyploids, greater height and stem-base diameter, and occupied a wider range of climatic conditions than conspecific native-range populations. We also tested whether the invasive populations had overcome genetic founder effects. We sampled a total of 80 populations in parts of the invaded range in China and the native range in North America for in situ measurements of plant height and stem-base diameter in the field and for population genetic and cytotype analyses. To examine climatic correlates, we augmented our field sampled data with occurrence records obtained from the Global Biodiversity Information Facility. All, except one, of the populations that we sampled in China occurred in a humid subtropical climate. In contrast, the North American populations occurred in humid continental, humid subtropical, and semi-arid climatic zones. All populations of S. canadensis in China were purely hexaploid while the North American populations were diploid, tetraploid, and hexaploid. The invasive hexaploids were significantly taller and had larger stem-base diameter than native hexaploids. Native hexaploids were significantly taller and had larger stem-base diameter than native diploids. Climatic correlate assessment found that invasive and native populations occupied different climatic envelopes, with invasive populations occurring in warmer and less seasonal climates than native populations. However, there was no significant correlation between ploidy level and climatic envelope of S. canadensis. Molecular phylogeography data suggest reduced genetic founder effects in the invaded range. Overall, these results suggest that polyploidy does not influence S. canadensis climatic tolerance.

Sampling and functional trait measurement

To identify occurrence locations of S. canadensis in China and North America, we conducted field surveys in summer and autumn for the period 2013-2016. We sampled a total of 2,051 S. canadensis individuals in 19 native and 61 invasive populations. The number of individuals sampled per population ranged from five to 33. The native range in North America was represented by 577 individuals while the invaded range in China was represented by 1474 individuals (Appendix 1). Any two sample populations were at least five kilometers apart. In each population, we sampled at random plants that occurred within a radius of 10 – 20 meters to minimize a potential bias due to environmental heterogeneity within a population (e.g. soil moisture and nutrients). To avoid sampling plants from the same genet within the same radius, we sampled plants that were at least two meters apart from each other.  . For each S. canadensis individual, we took in situ measurements of stem-base diameter and plant height (in cm). These functional traits are generally linked to ecological strategy axes of plants. As a plant developmental stage (i.e. phenophase) can influence the values of these traits, we took measurements only from mature plants that had set seeds and were therefore presumably not undergoing further development in height and stem diameter. We omitted leaf measurements because an individual S. canadensis plant can have hundreds of leaves (especially in the non-native range), and hence finding the largest leaf for measurement can be extremely difficult. In the native range, the ecological niches of S. canadensis and those of its congeners S. gigantea and S. altissima overlap and the three species can co-exist (Weber 1998, Benelli et al. 2019). However, S. canadensis prefers loose and drier soils than the congeners, and hence it occurs mostly near urban areas, roadsides, and railways, while the congeners occur mainly on riversides and embankments (Benelli et al. 2019). Therefore, within a locality, we sampled only the preferred habitats of S. canadensis to avoid accidental sampling of the congeners. Moreover, we conducted a test of phylogenetic relatedness among the populations that we sampled in the field to determine whether they all belonged to S. canadensis (see the section on population genetics and phylogenetic analyses below). All the sampled populations in China occurred in a humid subtropical climate except one that occurred in a subtropical highland climate. In contrast, the North American populations were sampled from humid continental, humid subtropical, and semi-arid climatic zones (Köppen–Geiger climate classification system)(Kottek et al. 2006).

Statistical analyses

To test whether S. canadensis plants from the invaded range had significantly greater height and stem - base diameter than S. canadensis plants from the native range, we fitted linear mixed-effect models. In the models, a trait value was treated as a dependent variable, while S. canadensis range (invaded versus native) was treated as a fixed-effect independent variable. Population identity of S. canadensis was treated as a random-effect factor and nested within range. To test whether polyploidy in general conferred higher mean trait values, we also fitted linear mixed-effect models to test whether the functional trait values differed between hexaploid and diploid individuals in the native range. Because only hexaploids were found in the invaded range, this test was not possible for invasive plants. In the test, ploidy level (polyploid versus diploid) was treated as a fixed-effect independent variable, while S. canadensis population identity was treated as a random-effect independent variable. We did additional analyses to test for the potential confounding effect of climate on trait expression by comparing traits of invasive plants with those of native plants from the same climatic zone only. As nearly all invasive plants were sampled in a humid subtropical climate, we used a subset of plants from this climatic zone only. We did separate comparisons for groups of plants that occurred in populations with mixed-ploidy and for groups of plants that occurred in populations with one ploidy type. Moreover, we compared traits of diploid plants with those of hexaploid plants that occurred in mixed-ploidy populations only to exclude the potential confounding effect of pure and mixed-ploidy populations occupying habitats with contrasting ecological conditions. In all the analyses, we fitted the models using maximum likelihood with the lme function in the nlme package (Pinheiro et al. 2007) in R v 3.5.2 (R Core Team 2018). We used likelihood ratio tests to assess the significance of each factor by comparing a model without the factor with a full model.

Assessing climatic tolerance

We described the climatic space occupied by S. canadensis with 19 variables with a 30 arcsec resolution (~ 1 km) derived from the WorldClim database (Hijmans et al. 2005). After standardization of the climatic variables (mean equal to zero and standard deviation equal to 1), we reduced the dimensionality of the climatic hyperspace to two axes (PC1 and PC2) based on a PCA (Broennimann et al. 2012). In addition to our 80 field-sampled records of S. canadensis in China and North America, we collected information on occurrences of S. canadensis provided by the Global Biodiversity Information Facility (GBIF) database (http://www.gbif.org/) in order to assess more comprehensively the climatic tolerance of the species. We retained only the GBIF occurrence records in the native (USA, Canada, and Mexico) and invaded (Australia, New Zealand, Japan and China) ranges where the species is most widespread. Occurrences flagged as invalid (e.g. in the oceans) or with doubtful coordinates were removed as well as those separated by less than 30 arc seconds to avoid spatial autocorrelation. We then extracted values of PC1 and PC2 for each of the 2,214 retained occurrence locations (i.e. our 80 field-sampled records and 2,134 GBIF records, which included 1,624 records in the native range and 510 records in the invaded range).

To test whether climatic space occupied by S. canadensis differed significantly between the invaded and native ranges, we calculated the kernel-smoothed density of occurrence of the two ranges based on the density of the different combinations of climatic conditions available to each range (Broennimann et al. 2012). The overlap between the climatic envelope of the two ranges was estimated with the Schoener’s D metric, which ranges from 0 (no envelope overlap) to 1 (identical envelopes) (Schoener 1968). We then tested for envelope equivalency and the less restrictive hypothesis of envelope similarity (comparing the observed values of envelope overlap to the 95^th percentile density of the simulated values) (Warren et al. 2008, who did not use ordination but species distribution model techniques). To achieve this, the D value between the two ranges was compared to a null distribution of D values computed between simulated envelopes built through randomization procedures (10,000 randomly selected ‘background’ points generated 100 times). This was computed with the R package ‘ecospat’ (Di Cola et al. 2017). As a D value does not allow a determination of whether an absence of envelope overlap results from the lack of overlap for one or two PCA axes, we used the function ‘niceOverPlot’ implemented in R for easier interpretation. To test whether S. canadensis inhabited a wider range of climatic conditions in its invaded range than in the native range, we performed an F-test of equality of variances (homoscedasticity) along the two PCA axes. We also used the same method to compare the climatic zones occupied by diploid and hexaploid populations of S. canadensis in its native range. Finally, we compared D values and envelope equivalencies and similarities between the following groups of S. canadensis populations: 1) North American versus Chinese populations that we had sampled in the field and with the same ploidy level, 2) North American versus Chinese populations that we had sampled in the field and without controlling for variation in ploidy level, 3) a global data set (i.e. the 80 populations that we had sampled in the field and GBIF records) of North America and Mexico grouped together versus China, and 4) a global data set of North America and Mexico grouped together versus all occurrence records in Australia, New Zealand, Japan, and China. Comparisons 1 and 2 tested the potential effect of variation in ploidy, while comparisons 3 and 4 tested whether our sampling was representative of the invaded and native ranges. For comparisons 3 and 4, it was not possible to control for variation in ploidy level because information on ploidy level was missing in the GBIF database.