Will climate change increase hybridization risk between potential plant invaders and their congeners in Europe?

Abstract Aim Interspecific hybridization can promote invasiveness of alien species. In many regions of the world, public and domestic gardens contain a huge pool of non‐native plants. Climate change may relax constraints on their naturalization and hence facilitate hybridization with related species in the resident flora. Here, we evaluate this possible increase in hybridization risk by predicting changes in the overlap of climatically suitable ranges between a set of garden plants and their congeners in the resident flora. Location Europe. Methods From the pool of alien garden plants, we selected those which (1) are not naturalized in Europe, but established outside their native range elsewhere in the world; (2) belong to a genus where interspecific hybridization has been previously reported; and (3) have congeners in the native and naturalized flora of Europe. For the resulting set of 34 alien ornamentals as well as for 173 of their European congeners, we fitted species distribution models and projected suitable ranges under the current climate and three future climate scenarios. Changes in range overlap between garden plants and congeners were then assessed by means of the true skill statistic. Results Projections suggest that under a warming climate, suitable ranges of garden plants will increase, on average, while those of their congeners will remain constant or shrink, at least under the more severe climate scenarios. The mean overlap in ranges among congeners of the two groups will decrease. Variation among genera is pronounced; however, and for some congeners, range overlap is predicted to increase significantly. Main conclusions Averaged across all modelled species, our results do not indicate that hybrids between potential future invaders and resident species will emerge more frequently in Europe when climate warms. These average trends do not preclude, however, that hybridization risk may considerably increase in particular genera.


| INTRODUCTION
Biological invasions are an important component of global environmental change and may have severe ecological as well as economic impacts (Bellard, Cassey, & Blackburn, 2016;Vilà et al., 2011). Owing to intensified trade and traffic, the global redistribution of species and their subsequent establishment outside their native range (=their naturalization) have considerably increased during the recent decades and are likely to further increase in the future (Seebens et al., 2015).
Pro-active management of such invasions is, however, hampered by the difficulty of predicting which species may become invasive and where. Such predictions are difficult because of the complex causes of invasions, which include biological traits of the invading species, biotic and abiotic characteristics of the recipient environment, and historical contingencies (Catford, Jansson, & Nilsson, 2009;Richardson & Pyšek, 2006). There are, however, a number of factors known to facilitate invasions such as early reproduction, rapid growth rate, efficient longdistance dispersal or specific trait profiles which are complementary to those of the resident biota (Buhk & Thielsch, 2015;Carboni et al., 2016;van Kleunen, Weber, & Fischer, 2010;Küster, Kühn, Bruelheide, & Klotz, 2008;Pyšek et al., 2015).
Apart from these factors, interspecific hybridization has been assumed to foster invasions since a seminal paper of Ellstrand and Schierenbeck (2000). Indeed, there are prominent examples of highly invasive hybrids. For instance, several species of the genus Tamarix have been introduced to North America during the 19th century.
Although all of these species have escaped cultivation, by far the most successful and widespread invader is the hybrid between T. ramosissima × T. chinensis (Gaskin & Kazmer, 2009;Gaskin & Schaal, 2002).
The possible reasons for hybrid success include increased phenotypic or genotypic variability, phenotypic novelty arising from transgressive segregation or adaptive introgression, and heterosis effects (Prentis, Wilson, Dormontt, Richardson, & Lowe, 2008). Heterosis effects may be maintained especially when hybridization is accompanied by allopolyploidization and/or a shift to apomictic reproduction, which sustain heterozygosity.
As species are transported around the world with increasing intensity, barriers to gene flow between once geographically separated species are reduced and new hybrids between introduced and resident species will probably emerge more frequently (Thomas, 2013).
For the British Isles, a recent overview has already demonstrated a rise in the number of hybrids during the last few decades (Stace, Preston, & Pearman, 2015). Apart from the risk that the new hybrids include particularly successful future invaders, rising hybridization rates also raise conservation concerns (Bohling, 2016). In particular, genetic introgression and outbreeding depression may severely threaten native species (Todesco et al., 2016), especially those that are rare and only exist in small populations (Bleeker, Schmitz, & Ristow, 2007).
Disregarding deliberate crossings (e.g., for horticultural reasons), the risk of hybridization between introduced and resident species will depend on the introduced species' ability to naturalize, that is to establish self-sustaining populations in the wild, because naturalization intensifies the spatial contact of the newcomers with their potential hybridization partners in the regional flora and hence increases mating opportunities. The likelihood of naturalization of an introduced species is mainly determined by propagule pressure (Simberloff, 2009) and the suitability of abiotic and biotic conditions (Pyšek et al., 2012;Shea & Chesson, 2002). Among the abiotic factors, climatic suitability has been repeatedly shown to play a Results: Projections suggest that under a warming climate, suitable ranges of garden plants will increase, on average, while those of their congeners will remain constant or shrink, at least under the more severe climate scenarios. The mean overlap in ranges among congeners of the two groups will decrease. Variation among genera is pronounced; however, and for some congeners, range overlap is predicted to increase significantly.

Main conclusions:
Averaged across all modelled species, our results do not indicate that hybrids between potential future invaders and resident species will emerge more frequently in Europe when climate warms. These average trends do not preclude, however, that hybridization risk may considerably increase in particular genera.

K E Y W O R D S
alien ornamental plants, climate change, interspecific hybridization, invasion biology, range overlap, species distribution models prominent role (e.g., Feng et al., 2016;Hayes & Barry, 2007;Thuiller et al., 2005). As a corollary, predicted climate change is also likely to alter the naturalization odds of introduced alien species and thus the likelihood that they hybridize with resident species (e.g., Bellard et al., 2013).
The alien flora of a region consists, first, of plant species that have already become naturalized or invasive. In addition, there is an often much larger group of alien species that have been introduced to a region and are grown there but have not escaped from cultivation yet. The latter group of species forms a massive pool of potential future additions to the regional wild flora. In many regions, this pool is dominated by non-native plants used for public and domestic gardening (Hulme et al., 2008;Niinemets & Penuelas, 2008;Pergl et al., 2016). In Europe, for example, more than 16,000 species from more than 200 families are currently in cultivation for ornamental purposes, with many of them being alien to Europe (Cullen, Knees, & Cubey, 2011). Some of these non-native garden plant species have already become naturalized or invasive elsewhere in the world (van Kleunen et al., 2015) and can hence be considered particularly likely to do so in Europe too (Williamson, 1999).
In a recent paper, Dullinger et al. (2016) showed that this latter group of "alien garden plants naturalized elsewhere" will benefit from a changing climate in Europe in as much as the area climatically suitable to them will increase. Given that climatic suitability is an important prerequisite to alien species' naturalization and that naturalization facilitates hybridization of introduced and resident species, the risk that new hybrids emerge may thus also be expected to increase in the future. The newly establishing garden plants may thereby hybridize with resident (i.e., native and already naturalized or even invasive) species (e.g., Ayres, Smith, Zaremba, Klohr, & Strong, 2004). However, a climate-driven modification of regional hybridization risk does not only depend on the naturalization odds of garden plants, but also on changes in climatically suitable ranges of their potential hybridization partners (Dehnen-Schmutz, 2011).
In other words, the changing spatial overlap in areas climatically suitable for alien garden plants and for their potential resident hybridization partners in the wild (both native and naturalized) flora will determine possible changes in the risk of hybridization between these two groups.
Here, we evaluated whether climate change may lead to an increase in this spatial overlap. We studied a group of 783 alien ornamental plants not yet naturalized in Europe, but established outside their native range elsewhere in the world, as identified in Dullinger et al. (2016). From this group of 783 species, we first selected all those belonging to genera with hybridization documented in the literature.
We then fitted species distribution models for this subset of non-native ornamentals as well as for all their congeners in the native and naturalized European flora. We restricted our analysis to congeners because hybridization risk is strongly linked to genetic distance (Mallet, 2005), and intergeneric hybrids are rare (Whitney, Ahern, Campbell, Albert, & King, 2010). Finally, we assessed to what extent the range matching between the selected garden plants and their congeners will increase under three different climate change scenarios.

| Species selection
Our initial pool of study species was the same as used by Dullinger et al. (2016). These authors aligned the European Garden Flora Here, we used a systematic web-based literature search to further narrow this group of candidate species to those particularly relevant in the context of both invasion and hybridization. We used all possible combinations of the following keywords in the Web-of-Science (http://apps.webofknowledge.com): #hybridization, #hybridisation, #invasion, #alien, #invasive species, #plant. The records were subsequently limited to the following categories: agriculture, biodiversity, conservation, ecology, environmental sciences, evolutionary biology and reproductive biology. We screened the abstracts of the 1,220 papers found and finally identified 66 plant genera that fulfil the following criteria: (1) interspecific hybridization has been documented and (2) they contain invasive species (even if these are not identical with the hybrids or if only intraspecific hybrids have so far been reported to be invasive, for example in Pyrus (Hardiman & Culley, 2010)). Twentythree of these genera were represented by at least one species in the list of Dullinger et al. (2016), of which 18 were also represented by at least one species (native and naturalized) in the flora of Europe (Tutin et al., 1964(Tutin et al., -1980. From these, we discarded the genera Rosa and Rubus because of taxonomic difficulties with a large number of apomictic species. As a result of these consecutive filtering steps, we ended up with 16 genera. These 16 genera contain 34 alien plants currently cultivated in Europe with the potential to escape into the wild (indicated by their naturalization in other continents) and at least one congeneric species in the native and naturalized flora of Europe which shares the same life form (assuming that only mating partners of the same life form are likely to produce viable hybrid offspring; see Tables S1, S3, S6). Most of these species are planted for ornamental purposes only, but some, like Chenopodium quinoa or several Eucalyptus spp., are also of commercial interest beyond horticulture. After a final screening in GBIF for those species with more than 50 occurrence records (see Table S2), the group of congeneric species within Europe contained 133 native and 40 alien naturalized spp (see Table S6).

| Species distribution data and climatic maps
Data on the world-wide distribution of the 34 alien garden plants and their 173 native and naturalized congeners were taken from GBIF.
All species lists were taxonomically harmonized using The Plant List (http://www.theplantlist.org). Multiple occurrences within 10′ × 10′ grid cells and clearly erroneous records, that is those in water bodies, were removed. We did not limit records to those from the native range because species are known to partly expand their realized climatic niches in the naturalization range (Dellinger et al., 2016;Early & Sax, 2014;Petitpierre et al., 2012). Quarter, BIO18-Precipitation of Warmest Quarter. All these variables are known to potentially influence species distributions (Root et al., 2003). All climatic variables were provided by wORLdcLim at a spatial resolution of 10 min.
Possible future climates in Europe were represented by three emission scenarios of the IPCC5-scenario family: the milder RCP2.6, the medium RCP4.5 and the severe RCP8.5 (IPCC, 2013). The respective monthly temperature and precipitation time series, already regionalized for Europe, were taken from the Cordex portal (http://cordexesg. dmi.dk/esgf-web-fe/live) and used to recalculate 10′ resolution maps of the above six bioclimatic variables for possible future climates of the 21st century. A 50-year average of the period 2050-2100 was then used as the climate of the future in model projections (see below). probabilistic output of the two ensemble models was aggregated to a weighted mean, with weights determined by their respective TSS scores. Similarly, binary outputs of each of the two ensemble projections were generated based on a threshold that maximizes the TSS score (Liu, Berry, Dawson, & Pearson, 2005;Liu, White, & Newell, 2013) and then aggregated to a conservative consensus ma; that is, 10′ resolution cells were only classed as climatically suitable to a species if both ensemble models agreed on the potential presence of the species in the cell. Both metrics were calculated for each possible species pair; that is, each of the 34 garden plants was combined with any of its congeners.

| Overlap of climatically suitable ranges
Overlap metrics were subsequently averaged per species of garden plant (i.e., the average range overlap of each garden plant species and all its congeners in the wild flora was computed), separately for each climate change scenario. These average overlaps were then compared among the current climate and each climate change scenario using linear mixed-effects models (LMMs). Each LMM used the 34 ratios of current-to-future climatic range overlaps as the response, which was regressed against a fixed intercept, that is we tested whether the mean of the logarithm of these ratios was significantly larger or smaller than 0. A random intercept for genus was estimated to account for the fact that some genera were represented by more than one species of garden plant.
All analyses were carried out in R (R Core Team, 2015) mainly using the packages RastER (Hijmans & van Etten, 2012)

| Geographic overlap of suitable ranges
Species distribution models for both the 34 alien garden plants and their 173 congeners in the native and naturalized European flora produced accurate projections in most cases (see Table S6).
True skill statistic scores suggest that the mean geographical overlap between the climatically suitable ranges of the 34 garden plants and their congeners will decrease under a warming climate (Figure 1a Table S4).
When overlap is measured as the number of 10′ × 10′ cells that are climatically suitable to both the garden plants and their congeners (i.e., the absolute size of their overlapping range, see Table S5), the results suggest that a warmer climate will not change the size of overlapping ranges in a statistically significant way in any of the scenarios (Figure 1b, see Table S4).
Looking at climatically suitable ranges of the 34 garden plant species and their 173 congeners separately indicates that these results are partly driven by opposite effects of climate change on the two species groups: while average range size (=number of suitable cells) is projected to increase for the garden plants (statistically significantly only for scenario RCP8.5, see Table S4), it will remain constant or even decrease for their congeners in the wild European flora, at least under the more severe scenarios (RCP4.5 and RCP8.5, Figure 2 and see Table   S4). These opposite trends apparently result in no net change in overlap or in a slight reduction depending on scenario and overlap measure used, but never in a significant increase in overlap.
These average trends mask strong differences among genera.

| DISCUSSION
Taken together, our results do not support the expectation that the area suitable to both the group of potential future invaders among European garden plants and their congeners in the resident flora of the continent will increase under a changing climate. Potential range overlap between these two groups of species will rather decrease under all warming scenarios. This is partly due to opposing trends in the size of climatically suitable ranges among the two groups: while potential invaders on average expand their suitable ranges, those of resident congeners remain constant or shrink, at least under moderate and severe warming. However, there is pronounced variation among the different species pairs and for some of them the predicted increase in range overlap is significant, suggesting that the risk of hybridization between them will also increase.
Climate change has already allowed many alien species to expand their non-native ranges (e.g., IPCC, 2014). For ornamental plants, the plants used in this study are also mostly native to warm regions and hence their potentially suitable ranges in Europe tend to increase, on average, despite pronounced idiosyncratic differences. Although the pool of their European congeners contains many warm-adapted species too (e.g., most species from the genera Euphorbia and Tamarix), it also includes a considerable number of montane or even alpine species (e.g., from the genera Linaria, Rhododendron, Senecio and Viola).
For montane species, climatically suitable ranges are particularly likely to shrink under climate warming (Engler et al., 2011;Thuiller et al., 2014). The share of montane species is thus probably a factor restricting range increases of congeners in the more severe climate scenarios.
We emphasize that our estimate of changing range overlaps does not include a temporal dimension. Real changes in overlap of species distribution over the 21st century may actually deviate from those projected here. On the one hand, wild populations of species (both native and naturalized) will likely lag behind the changing climate due to dispersal and migration constraints (e.g., Corlett & Westcott, 2013;Dullinger et al., 2015). These constraints are less relevant or even irrelevant for ornamental plants in horticultural trade. Actually, garden plants may even "overtake" climate change when regional demand of gardeners anticipates future climatic alterations (Bradley et al., 2012).
On the other hand, remnant populations of species in the wild may still occupy an area long after the average climate has become unsuitable to them (Eriksson, 2000). Actual range overlap over the next decades will hence not only be a function of changes in suitable ranges, but will be co-determined by the behaviour of gardeners and by migration lags and extinction debts of wild populations (Dullinger et al., 2012). Thus, we may expect that our SDM-based projections will underestimate real overlap near the wild species' trailing edges (because of delayed extinctions), but overestimate it near the wild species' leading edges (because of lagged migration).
An average decrease in range overlap among all the species pairs tested here does not necessarily imply a general decrease in hybridization risk from invasive plants in Europe. First, we deliberately restricted our approach to hybridization among potential future invaders and resident species but did not consider the possible emergence of hybrids within the resident (i.e., native and already naturalized or even invasive) species. Among the latter, several hybrids come from genera well-known to hybridize such as Fallopia (Parepa, Fischer, Krebs, & Bossdorf, 2014) or Epilobium (Gregor et al., 2013). For an exhaustive evaluation of climate-driven changes in hybridization risk of non-native plants, these species would have to be included into the models. Second, the probability of hybridization risk will likely vary widely among the species pairs included in this study. Successful establishment of allopolyploid hybrids, for example, depends on plant traits (Mallet, 2007). In addition, the genetic distance between species certainly differs a lot among the pairs studied and hence also the likelihood that reproductive barriers break down (Mallet, 2005). A more precise evaluation of hybridization risk under climate warming would therefore have to weight changing range overlaps by the likelihood that particular species pairs hybridize at all-and, in an additional step, by the probability that a particularly successful invader emerges from such hybridization (e.g., Abbott et al., 2009;Hovick & Whitney, 2014).
Such weighting might significantly modify expected changes in hybridization as individual species pairs with increasing range overlap are to be found in almost all genera. Although data for reliable estimation of these weights are lacking, we emphasize that among the genera with increasing average range overlaps in at least some scenarios, species in Solidago and Rhododendron have already produced invasive hybrids in Europe (Abbott et al., 2009;Erfmeier, Tsaliki, Ross, & Bruelheide, 2011;Karpaviciene & Radusiene, 2016) and may hence be particularly likely to do so in the future again. In addition, among the genera which were both identified to have produced invasive hybrids in the metaanalysis of Hovick and Whitney (2014) (Bleeker et al., 2007), gene swamping (Todesco et al., 2016) or pollen competition (Arceo-Gomez & Ashman, 2016). Among the genera included in this study, introgressive hybridization has been documented in several cases (e.g., Tamarix (Gaskin & Kazmer, 2009), Rhododendron (Stace et al., 2015), Viola (Stace et al., 2015)). Conversely, Bleeker et al. (2007) (Bohling, 2016), future escape and expansion of ornamental plants into the range of these endemics may actually put additional pressure on them, beyond the challenges they face under a warming climate. Most of these species are not included in our study as their distribution is not represented well enough in GBIF, but this issue certainly warrants further investigation.
Finally, as a last caveat, we note that the models this study is based on were fitted using data taken from GBIF. This source combines the advantage of a global coverage, and hence the possibility to fit niches of species comprehensively, with the disadvantage of the errors and biases implicit to this database (Meyer, Weigelt, & Kreft, 2016).
Uncertainties in species distribution estimates and models resulting from these caveats have a clear geographical bias and are least pronounced in the well represented regions of Europe, North-and Central America, and Australia (Meyer et al., 2016). The majority of the ornamental plants and all congeners modelled here come from these areas, and we hence assume that data problems are of limited importance for them. Several of the ornamental plants are native to temperate Asia and Africa, however, and these regions have notoriously low data coverage. The most likely consequence of this low coverage is an underestimate of these species' niches and hence of their potential distribution in Europe as well as their overlap with native and already naturalized congeners. Such underestimation may have been reinforced by the restrictive rules of our consensus projections. As a result, range overlap estimates computed here are probably conservative. We do not, however, think that these data problems affect our main result, namely that the average potential range overlap between ornamental plants and congeners does not increase under a warming climate. This is because predicted trends for species of Asian and African origin are similar to those of the remaining species (see Fig. S1 and Fig. S2 respectively).

| CONCLUSION
Climate warming will potentially increase the area suitable for the naturalization of many non-native ornamental plants in Europe , but the mean geographical overlap of climatic ranges between the selection of ornamentals and their native and naturalized congeners modelled here is unlikely to increase in the future. Thus, the average risk that garden plants and their wild congeners in the European flora will hybridize does not appear to rise when climate warms. We emphasize, however, that suitable range overlaps do increase for many individual congener pairs and that the pair-specific likelihood of successful hybrid establishment is unknown. A decreasing average range overlap does not, therefore, preclude increasing invasion risk from hybrids between particular species pairs.