Reproductive strategy of a temperate canopy tree Tilia cordata Mill. (Malvaceae) is related to temperature during flowering and density of recent recruits

Facultative clonality is extremely common in plants, but the relative emphasis on sexual versus asexual reproduction varies both between and within species, which in turn may influence individual fitness and population persistence. Tilia cordata is a temperate, entomophilous canopy tree that is partially clonal. Favourably warm climatic conditions have been linked with successful sexual reproduction in the species with clonality being suggested as the reason for population persistence in colder periods. Despite this the extent, character and structure of asexual reproduction in the species have never been described, nor has its relationship with climate. Fine-scale spatial genetic structure was assessed in 23 stands across a latitudinal gradient. The proportion of individuals that are of clonal origin has a wide range with a mean of ~43%. Genetic diversity is high, with even mostly clonal stand possessing several distinct genotypes. A beta regression model shows that historic summer temperatures and density of recent recruits are predictors of the proportion of clonal recruitment. Clonal reproduction is less important in stands that experience higher temperatures during flowering while stands with more saplings have more clones. Additional factors likely affect the balance between the two reproductive modes. The climatic relationship suggests a trend towards a higher proportion of recruitment from seed in a warming climate, although factors such as herbivory may prevent this.


Introduction
Although obligate asexual reproduction is rare in angiosperms, facultative clonality is extremely common (Barrett 2015;Eckert 2002;Harper 1977), particularly in perennial plants (Nuortila et al. 2002;Holsinger 2000). Vegetative propagation is the most common form of clonality with an estimated ~80% of all angiosperm lineages possessing some means of reproducing in this manner (Klimeš et al. 1997). Compared to exclusively sexual reproduction, a mixed reproductive strategy provides several ecological benefits. These include an increased capacity for resource acquisition in patchy environments (light, nutrients, water), lowered chance of death for a particular genotype due to spread mortality risks, and the possibility of physiological integration and spatial division of labour (e.g. sharing of root systems) amongst shoots (Barrett 2015). Clonal reproduction may also provide a buffer against complete reproductive failure in periods of poor weather or when conditions are otherwise unsuitable for sexual reproduction (Barrett 2015).
However, over longer time scales, clonality can have negative consequences on individual fitness and population duration. It reduces genotypic diversity by definition, and if species are self-incompatible and the spatial arrangement of vegetative growth is aggregated, outcrossing success can also be reduced (Vallejo-Marín et al. 2010;Charpentier 2001). This can affect the viability of small populations due to lower sexual fecundity as a result of intraclone incompatibility (Honnay and Bossuyt 2005;Nuortila et al. 2002). This process can ultimately lead to a complete failure of sexual reproduction and if any particular clonal lineage has Communicated by K. Heer a competitive advantage, monoclonal stands (Halkett et al. 2005;Honnay and Bossuyt 2005). Reduced genotypic diversity and a lack of sexual reproduction can affect future adaptive potential by lowering the efficiency of natural selection or increase the rate at which deleterious alleles can accumulate, further decreasing reproductive capacity. Either of these factors could affect the likelihood of long-term population persistence (Holsinger 2000); hence, the necessity of understanding both how common clonal reproduction is as well as its character and spatial arrangement.
The relative emphasis on sexual versus asexual reproduction differs not just between but also within species (Silvertown 2008), with different contexts favouring one reproductive mode over the other. For instance, in widely distributed species, clonality is often more prevalent in marginal populations which are subject to less favourable environmental conditions (Brzosko et al. 2013). If these reduce the success rate of sexual reproduction or cause it to fail entirely, vegetative propagation can contribute to population persistence due to the longevity of clonal individuals and the ability to continue recruiting (Barrett 2015;de Witte and Stöcklin 2010). Specific factors that can shift the balance between sexual and asexual reproduction include disturbance regime (Morris et al. 2004), water depth in aquatic species (Li et al. 2018), individual density (Rautiainen et al. 2004) and sex ratios in dioecious species (Dering et al. 2016). Climatic factors such as water availability have also been shown experimentally to shift the balance between the two modes by increasing the relative cost of sex (Wang et al. 2018). Broadly speaking, species that exist in environmental extremes such as arctic regions, aquatic or shady habitats and nutrient poor conditions show a tendency towards clonality. This suggests a correlation between clonality and habitat quality, amongst other aspects of environmental variation (Ye et al. 2014(Ye et al. , 2016van Groenendael et al. 1996).
The climate is rapidly changing (IPCC 2021); this will in turn have an impact on the distribution of species; an understanding of the balance of reproductive modes in a species can be an important predictor of migration potential given the relative effectiveness of seed versus vegetative dispersal (Morris et al. 2014;Silvertown 2008). This makes an understanding of clonality potentially important information in the generation of species distribution models used to forecast responses to climatic change (as these can benefit from the inclusion of life-history traits, e.g. Matthews et al. 2011). Similarly, the impact of some plant pathogens is expected to increase (e.g. Phytophthora spp.) as a result of more favourable environmental conditions or an increased susceptibility of host species due to factors such as drought stress (Sturrock et al. 2011). As biotic stress tolerance generally and pathogen resistance specifically are linked to genetic diversity (or the ability to generate novel genotypes via recombination), an understanding of clonality may be part of mitigating these impacts (Honnay and Bossuyt 2005). Finally, where genetic variation is at risk and the need for conservation has been realised, the knowledge of this aspect of an organism's life history allows appropriate management measures that best capture and maintain the species' genetic variation to be chosen (Namroud et al. 2005).
The difficulty of examining reproductive balance and determining its correlates varies between taxa. Communitylevel studies that look at prevalence of clonality in a particular habitat have demonstrated environmental correlates of the same (e.g. Ye et al. 2014Ye et al. , 2016Silvertown 2008). The difficulty with such an approach lies in removing phylogenetic constraints (van Groenendael et al. 1996), which brings to the fore the importance of examining withinspecies clonal variation as a way of elucidating the role of environmental factors in inducing shifts in reproductive mode (Morris et al. 2004). The previously cited examples of shrubs and forbs show that some species can be manipulated experimentally to assess directly how a particular abiotic or biotic variable affects clonal growth. For trees however, their slow growth and size can make such experiments unfeasible, and studies must be undertaken using natural populations. A wide variety of tree species undertake some form of clonal propagation, including many Quercus (Fagaceae), Populus (Salicaceae), Fagus (Fagaceae) and Tilia species (Malvaceae) (Logan et al. 2019;Evans and Morris 2016;Guarino et al. 2015;Logan et al. 2015;Morris et al. 2014;Lone et al. 2011;Mock et al. 2008;Alfonso-Corrado et al. 2005;Bakker et al. 2001;Montalvo et al. 1997;Pigott 1991). Trees have been shown to exhibit within-species variation in the balance between sex and vegetative propagation, just as other plants do. For example, differences in clonal incidence between stands of oak have been linked with distinct disturbance regimes (Valbuena-Carabaña and Gil 2017). Previous work describes or tests for differences in the incidence of clonality between locations in distinct environmental contexts (low, high altitude) or broad categories of habitats (marginal, central) (Logan et al. 2019;Morris et al. 2004Morris et al. , 2014Persson et al. 2004). To our knowledge, however, no explicit link between climate and clonality in temperate canopy trees has been demonstrated; a direct comparison of climatic variables and reproductive strategy is lacking. This represents a significant gap in the literature given the role of trees in dominating and shaping their community.
Tilia cordata Mill. (small leaved lime) is a species with a potential connection between climate and reproductive balance. It is an entomophilous canopy tree with a widespread but patchy distribution (Fig. 1). An effective vegetative propagator, it produces clonal individuals ('ramets', sensu Harper 1977; cf. 'genets', the clonal group as a whole) in a variety of ways. Shoots from the root collar can develop into multiple stems and aboveground connections can disappear over time, leading to separate individuals. These basal shoots can also layer themselves, while fallen trees can produce multiple vertical shoots along their length, which ultimately develop into new ramets. The capacity for vegetative propagation does not seem to reduce with stem age (Pigott 1991). By contrast, successful sexual reproduction in the species seems to rely on critical temperatures being reached during both pollination and seed development (Pigott 1981;Pigott and Huntley 1981,).In the UK, flowering is typically early to mid-July in the south and mid to late July in the north (Pigott 1981(Pigott , 2012Pigott and Huntley 1981), with a 50-day difference on flowering times across the full latitudinal range (Pigott 2012). The current extent of its range was established during the Holocene climatic optimum when summer temperatures were 1-2 °C higher than currently (Birks 1989;Davis et al. 2003), and its fertility was presumably not an issue. The subsequent temperature decline means that northern marginal populations rarely set fertile seed except during the warmest summers, due to either a failure of pollination (Pigott and Huntley 1981) or of fruit development (Pigott 1981). This is in contrast to more southern populations which frequently produce viable fruits (Pigott 1991). In concert with other potential factors such as herbivory (Pigott 1991), this ultimately leads to infrequent sexual recruitment, which may promote vegetative reproduction, or at least increase its relative importance (Silvertown 2008). As a result, clonal growth has been suggested as the reason for population persistence in marginal locations such as the north-west of England ( Fig. 1) (Pigott and Huntley 1978). Recent work in Denmark, also close to the northern limit of the species, found clonal reproduction in eight of the nine study sites and all four of those that were exhaustively sampled, suggesting the clones may simply have not been detected at the ninth site (Erichsen et al. 2019). In comparison, sites in central range locations can have little to no clonal growth at all (Logan et al. 2019); however, the clonal status of the most recent recruits in northern populations has yet to be investigated so it is unclear if this may have changed over recent years.
Although it is clear that tree species modify their reproductive strategy based on environmental differences between populations, the relationship between clonality and relevant climatic variables has not been examined explicitly. T. cordata is an ideal model organism for testing this, given a potential mechanism in temperature-linked fertility and already observed differences between northern and southern populations in terms of clonality levels. Furthermore, its current distribution in the UK is largely the product of natural processes rather than forestry practices, being rarely planted (Peterken and Mountford 2017;Rackham 2003), in contrast with other species (Rodwell 1991). This presents a natural experiment that would allow a potential clonalityclimate link to be identified and quantified. Understanding this relationship would add to our knowledge of how environmental constraints shape reproductive strategy in temperate canopy tree species, which will be an important part of assessing the evolutionary potential of species in a climate that is selectively dynamic. Here, then, we assess the levels of clonality in semi-natural populations of a canopydominant tree across a climatic gradient and in doing so test the hypothesis that reproductive balance is related to temperatures during flowering. Given current climatic trends, with increasingly warmer summers (IPCC 2021), the clonal status of the most recent recruits will also be established, as they may demonstrate changing strategies; this has not yet been assessed in T. cordata. Other factors have been linked with the frequency of asexual reproduction such as individual density and disturbance, and therefore, their relationship with clonality here will be assessed (Morris et al. 2014;Weed and Schwarzländer 2014;Valbuena-Carabaña and Gil 2013). Similarly, the consequences of clonality for population persistence depend on its character, such as the diversity  Table 1 for key to numbers and exact location); the dashed line indicates the approximate location of the northern range edge in the UK, based on Pigott (1991) of clonal genotypes and their spatial arrangement (Vallejo-Marín et al. 2010;Honnay and Bossuyt 2005), and so these will also be determined. Taken together, a direct comparison of climate and clonality, the addition of demographic comparisons between sites and a spatially explicit sampling scheme, represents a significant advance on previous such studies in T. cordata (e.g. Logan et al. 2019).

Sampling method and location
Eighteen semi-natural ancient woodlands (ASNW) containing populations of Tilia cordata were sampled. These areas were spread across a 274 km east-west and 376 km northsouth gradient, which represents a majority of the native range of the species within the UK, encompassing a spread of climatic conditions likely to be relevant to the balance between sex and clonal recruitment (Fig. 1A, Table 1). The most northern populations here are also close to the species' latitudinal range edge in Western Europe ( Fig. 1B) (Pigott 1991). Populations sympatric with the other UK native Tilia species (T. platyphyllos Scop., large-leaved lime) were not sampled to reduce the possibility of hybridisation as a confounding variable, particularly as both species are similar ecologically (Barker et al. 2015;Logan et al. 2015).
In each location, areas where T. cordata was a canopydominant were identified and a single 30 × 30 m quadrat was placed using random number generation. Leaf material was collected from all T. cordata individuals where possible and dried in silica gel (representative voucher specimens from each population were submitted to LIV and MANCH). To allow for an assessment of the spatial arrangement of clonality, the location of each tree was mapped using measuring tapes (Arnaud-Haond et al. 2007a). Trees often possessed multiple stems due to the sites' history of coppicing as a management strategy.
Therefore, only stems not connected aboveground and separated by a distance of more than 0.2 m were considered distinct individuals. Due to the history of coppicing as a management strategy in these populations, and the link between clonality and disturbance generally or this practice specifically, any stem part of an obvious coppice stool was noted as such. As stools gradually become separate individuals (Del Tredici 2001), the proportion of the quadrat composed of visible coppice stools should therefore serve as a proxy for time elapsed since the stand was last coppiced. This allows for an estimation of the relationship between disturbance regime, or at least that part of it caused by forestry practices, and clonal growth.
Similarly, if clonality is linked with climate, then changes in the origin of new recruits may have occurred over time due to global warming. It is important then to understand the origins of any younger individuals within quadrats, which has not been done yet in this species. Therefore, demography data was recorded. Stems were classified as adult or juvenile based on trunk diameter at 1.37 m (DBH), and overall height. Adult individuals were those with a DBH > 0.1 m, while juvenile individuals (both 'seedlings' and saplings) had an overall height of < 1.37 m and/or a DBH < 0.1 m (respectively). Occasionally, leaf material was inaccessible due to its height, and in these instances, the presence and position of these individuals were noted regardless (64 n). A total of 663 trees were sampled across all locations (Table 1).

DNA extraction & genotyping
Sample DNA was extracted using the 'crude extract' procedure described in a KAPA3G Plant PCR Kit (KAPA Biosystems, London, UK). Briefly, a 6.35-mm circular leaf punch was taken from each sample and placed in 125 μl of extraction buffer (50 mM Tris-HCl, 0.1 mM EDTA•Na 2 , 2% v/v ß-mercaptoethanol, 1 mM TCEP) before being heated at 95 °C for 5 m. This crude extract was used as the template for all genotyping.
Individuals were genotyped at 10 microsatellite loci (Table S1) using four multiplex PCR following Phuekvilai and Wolff (2013), with minor modifications. Loci were initially developed for T. platyphyllos and the selection here is based on cross-amplification success and polymorphism in T. cordata. Amplification was carried out in 10 μl total volume containing 1× Bioline MyTaq Plant-PCR Mix (0.2 mM each dNTP, 1.5 mM Mg 2+ ; Bioline Reagents Ltd., London, UK), 0.1-0.2 μM of each primer (Table S1), 0.5 mM TCEP and 1 μl of template. Forward primers were tagged with the fluorescent dyes NED, VIC (Applied Biosystems, Warrington, UK) or 6-FAM (Sigma-Aldrich, Dorset, UK). Dye colour and primer concentration can be found in Table S1. Thermal cycler conditions were as described previously (Phuekvilai and Wolff 2013).
PCR product size was determined using capillary electrophoresis on an AB3500 Genetic Analyser (Applied Biosystems, Warrington, UK). Amplicons were diluted 1:10 with nuclease-free water. One microliter of this dilution was mixed with 8.9 μl Hi-Di formamide and 0.1 μl LIZ500 size standard before running. Allele peaks were called automatically in GeneMapper 5.0 (Applied Biosystems, Warrington, UK) and then checked manually for errors. To reduce the potential for human error in the binning process, the software TANDEM, version 1.07 (Matschiner and Salzburger 2009) was utilised to produce integer allele sizes from the fractional sizing data produced by GeneMapper.

Population genetic parameters
For comparison with other studies, typical population genetic parameters were calculated using the package hierfstat, version 0.04-22 (Goudet 2005) in the R software package, version 3.4.2 (R Core Team 2017); all other analysis was also undertaken in R. Genetic diversity of quadrats was demonstrated by calculating the percentage of polymorphic loci, mean number of alleles and observed and expected heterozygosities. In addition to this, differences in the number of individuals in quadrats meant that simple counts of alleles are not directly comparable, and therefore, rarefied allelic richness (n = 13) was also calculated (El Mousadik and Petit 1996). Finally, the inbreeding coefficient F IS was calculated as an estimate of individual relatedness within the quadrats, and its 95% confidence interval estimated by bootstrapping (999 replicates).

Clonal identification
Identical multi-locus genotypes (MLG) were assumed members of the same clone. This identification was confirmed by assessing the probability that identical genotypes would be encountered a second time as a result of chance recombination events (p sex ), based on sampled allele frequencies (Parks and Werth 1993). This was calculated using the R package poppr, version 2.5.0 (Kamvar et al. 2014(Kamvar et al. , 2015. Where p sex was below 0.01 (Arnaud-Haond et al. 2007b), replicate genotypes were confirmed as the product of vegetative propagation.
The corollary to this is that individuals with distinct genotypes may actually be members of the same clone. Both somatic mutations and scoring errors introduced during genotyping can generate genetic differences in otherwise identical individuals (Meirmans and Van Tienderen 2004;Douhovnikoff and Dodd 2003;Klekowski 2003). The presence of these processes was inferred by producing histograms of inter-individual genetic distance as measured by the number of shared alleles (Arnaud-Haond et al. 2007a). These plots showed bimodal distributions, with a second peak at small but non-zero distances, which suggests that at least one of these processes has occurred (Arnaud-Haond et al. 2007a;Rozenfeld et al. 2007). Therefore, any pairwise comparison between individuals that differed by only one allele was considered to indicate membership of the same clonal group (Schnittler and Eusemann 2010). This clonal identity was confirmed as before by calculating p sex , but with the distinct locus temporarily removed from the data. Values of 0.01 or less were considered part of the same clonal multilocus lineage (MLL).

Incidence and character of clonality
The relative contribution of sexual versus asexual reproduction was assessed using a proportional measure of genotypic richness, R: where G is the number of distinct genotypes/lineages and N the total number of individuals. This varies from 0 when all sampled individuals are identical to 1 when all observed genotypes are distinct (Dorken and Eckert 2001). It is preferable to the often-used measure of proportion distinguishable (G/N), which can never reach zero even in populations composed of a single clone (Arnaud-Haond et al. 2007a). Summary statistics (mean and range) describing the number of ramets (N R ) within each clonal group were calculated. To describe how equitable clonal reproduction in T. cordata is, the complement of Simpson's index (1 -D, often called D*) was calculated (Arnaud-Haond et al. 2007a). To aid interpretation of this, a stacked bar plot showing the proportion of each quadrat occupied by each genotype was produced. This illustrates whether many lineages reproduce vegetatively, or if the production of clones is dominated by a small number of genotypes.
The spatial arrangement of clonal genotypes was described by mapping individuals and denoting clonal status. Due to the number of samples, the spatial size of multi-locus lineage (MLL) was quantified by calculating the maximum linear distance between clonal individuals (D max ) (Jarni et al. 2015). The degree of aggregation exhibited by each MLL was described using the clonal dominance index D c (Ohsako 2010). This is calculated for each group with ≥ 3 members: where N S is the group size (as number of ramets), and N T is the total number of individuals of any genotype occurring within the spatial range of particular MLL (here, a convex hull described by the outermost members of the lineage in question). This has been used to describe whether clonal groups are aggregated or intermingled with other genets (Dering et al. 2015;Chenault et al. 2011;Ohsako 2010). If MLL totally exclude other individuals (N S = N T ), then it reaches it maximum value of 1, while the value for more intermingled groups will approach 0.

Relationship of levels of clonality with climatic or demographic variables
A regression model was generated to test the hypothesis of an association between clonal reproduction and climate, and to explore whether other factors previously linked with clonality were involved. The proportional measure of genotypic richness R was used as the response variable and summaries of recent local climate during flowering were used as predictors. The WorldClim2 data set provided  Kilkenny and Galloway 2008;Herrera 1995;Abrol and Kapil 1986). Since T. cordata is primarily entomophilous (Pigott 1991;Anderson 1976), they could affect the success of sexual reproduction and therefore may be relevant when considering reproductive balance. As stands had differing numbers of trees and different demographic profiles, the density of adult and juvenile stems per square metre and the proportion visible coppice that were established trees were also included as predictors to account for any variation in clonal proportion these might generate (Cristóbal et al. 2014;Weed and Schwarzländer 2014;Valbuena-Carabaña and Gil 2013;Johansson and Lundh 1988). The response variable is bounded to the unit interval [0, 1], and so a beta regression model was used to avoid prediction of nonsensical values (i.e. > 1, < 0) (Simas et al. 2010;Ferrari and Cribari-Neto 2004). Calculations were performed using the R package betareg, version 3.1.0 (Cribari-Neto and Zeileis 2010). Beta regression also shares similarities with generalised linear models, in particular the use of a link function between the response and predictors. This provides additional flexibility, accounting for factors such as skew within the response (Ferrari and Cribari-Neto 2004).
Since our goal was partly descriptive and exploratory concerning other potential factors involved in generating reproductive balance, rather than predictive, model selection techniques were avoided (Mac Nally et al. 2018). A complementary log-log link function log e (-log e [1μ]) for the mean parameter μ was found to improve model fit over a logit link, log e (μ / [1μ]), significantly increasing log-likelihood based on a Wald test (Χ 2 = 0.09, df. = 1, p < 0.001) and raising Akaike information criterion (AIC) markedly (Canterle and Bayer 2017;Akaike 1974).
Pairwise Pearson's correlation coefficients were used to determine whether any predictors were collinear. Mean minimum, average and maximum daily temperatures were strongly correlated with each other as well as mean rainfall (r > 0.7), and so, all bar mean daily maximum temperatures were removed (Pigott and Huntley 1981). The two remaining climatic variables, water vapour pressure and solar radiation, were also collinear. Therefore, only solar radiation was retained due to ease of interpretation; light levels are commonly linked with pollinator behaviour (see above). Similarly, the density of sampled stands was strongly correlated (r > 0.7) with demographic variables such as proportional measures of age. Since the density of adult and juvenile trees was uncorrelated, these were retained.
This was validated using typical approaches. To ensure that the assumption of homoscedasticity was not violated; residuals were examined by plotting against fitted values. Since sampled populations were clustered in geographic locations due to the nature of T. cordata's scattered native distribution (Pigott 1991); independence of observations was assessed by using a Moran's I test to determine whether residuals exhibited spatial autocorrelation. Finally, the presence of influential observations was assessed by calculating Cook's distance for each data point.

Population genetic parameters
Levels of genetic diversity were high across all quadrats (Table 2). In all but one instance (site 10), 100% of loci were polymorphic. Mean number of alleles and the rarefied equivalent were broadly similar and do not appear to vary with latitude. Heterozygosity was fairly high and again similar between quadrats. The inbreeding coefficient F IS varied, but only three had a 95% confidence interval (CI) that did not overlap zero (sites 3, 13 and 18); here, the values were negative, indicating that observed heterozygosity was lower than expected.

Clonal identification
The final number of distinct multi-locus lineages (MLL) observed was 370, for a total proportion 44.2% clones overall. All p sex values for single re-encounters of the repeated genotypes were below the chosen alpha of 0.01, confirming initial clonal identification. Of the final set of repeated MLL, 32 had genotypes that differed from the other members of their clonal lineage by only one allele. After temporarily removing this locus and recalculating p sex , all values were below the original threshold of 0.01, confirming this additional clonal identification.

Clonal incidence and diversity
Clonal reproduction occurred at all sites, at variable levels. The proportion of non-clonal individuals R was typically over half of all trees sampled (mean = 0.57, standard deviation = 0.20). R ranged from 0.17 in quadrat #5 to 0.96 in Table 2 Samples and some of their observed population genetic parameters The sites are given in order from north to south, as elsewhere. Columns two through six represent the genetic diversity within each sample: A is the mean number of alleles across all loci; % P, percentage of polymorphic loci within a sample; RS, rarefied allelic richness (n = 13); H obs and H exp , observed and expected heterozygosity respectively; F IS , the mean inbreeding coefficient across loci; finally, the estimated 95% confidence interval (CI) for this mean, calculated by bootstrapping within samples (999 replicates quadrat #17 (Fig. 2, Table 3). The number of ramets (N R ) within each genet was typically low, with an overall mean of 3.34, and only one quadrat (#5) had a clonal group with more than ten members. Both the complement of Simpson's index show D* and stacked bar plot of genotypic proportions show that clonal reproduction is not usually dominated by any particular lineage (Fig. 3, Table 3), but evenness does vary between locations. Despite this, even those locations with large MLL and low genotypic richness such as quadrat #5 and #7 have reasonably high levels of diversity (D* = 0.84 and D* = 0.74, respectively), with several clonal genotypes observed. Spatial data indicates that clones are aggregated in typically exclusive groups, but that these groups are not large (Fig. 4). The mean maximum distance between members of the same clonal lineage (D max ) was low at 3.92 m, with an overall range of 0.33-21.6 m ( Table 3). Only seven sites had observed D max values greater than 10 m (e.g. Fig. 4A, B), with the least clonal sites having lower D max (e.g. Fig. 4C, D, E). The aggregation of these clonal groups was described by the clonal dominance index D c , which indicated that MLL with more than two members very rarely contained other genotypes within a convex hull described by their outermost members (mean D c = 0.98). Even quadrats such as #5 (Fig. 4A) and #7, with large distances between members of the same MLL and high numbers of ramets in clonal groups, have a mean D c of 1, showing no intermingling of genets.

Relationship of clonality with climate and other factors
A beta regression model showed that only some of the examined factors are significant predictors of the proportion of clonality observed (n = 18, pseudo-R 2 = 0.66; Table 4). Namely, the density of juvenile stems in a quadrat and the mean maximum daily temperature in July for the period 1970-2000 are significant predictors of the incidence of clonality (Fig. 5). Temperature explained slightly more variance in R than the density of juveniles (Table 4). Model validation showed reasonable fit and no violation of assumptions. The lack of obvious patterns when residuals were plotted against fitted values confirmed homoscedasticity (Fig. S1A); geographically clustered sample populations did not violate independence of observations based on a Moran's I test of the residuals, which showed no spatial autocorrelation (I = −0.05, p = 0.91). Finally, no calculated Cook's distance was above 0.5, showing that no data points had undue leverage (Fig. S1B).

Discussion
Despite tree species modifying their reproductive strategy between populations based on environmental differences between locations, an explicit comparison between a climatic variable and levels of clonality has not been made previously. Therefore, we assessed levels of clonality in seminatural populations of a temperate canopy-dominant tree, Tilia cordata, across a climatic gradient and compared these observations with historic temperatures during the critical period of flowering, amongst other factors. T. cordata was found to be partially clonal throughout its UK range and that current clonal incidence is correlated with historic July temperatures. These findings are supported by previous studies that suggest T. cordata is more clonal in marginal populations and that recruitment from seed is likely to be linked with temperature (Logan et al. 2019;Pigott 1991;Pigott and Huntley 1980). It contrasts with previous reports based on morphological observations which were identified northern populations as exclusively clonal (Pigott and Huntley 1981); recent recruits derived from seed were observed in one of the most northern populations (#2). Our results fit with other studies that suggest tree species modify their reproductive strategy between populations depending on environmental conditions. Such factors include altitudinal differences (Morris et al. 2004), marginal/central populations (Persson et al. 2004) and disturbance regime Gil 2013, 2017;Morris et al. 2014).
While almost all populations exhibit both clonal and sexual reproduction, populations at more northerly latitudes with cooler summers possess a greater proportion As a plausible mechanistic link between temperature during flowering and successful production of fertile seed has already been demonstrated experimentally (Pigott and Huntley 1981), this relationship is likely causative. Since other climatic variables were removed as predictors due to collinearity, the role of factors such as precipitation cannot be eliminated. Similarly, the exact nature of the link, whether the critical variable is minimum, maximum or mean temperature during flowering, requires further investigation. Since month-to-month temperatures are naturally correlated, other potential candidates for predictors of clonality include August temperatures, which are important for seed development (Pigott 1981).
The balance of reproductive modes is not predictable by wind speed or solar radiation during flowering. As T. cordata is primarily entomophilous (Fromm 2001;Pigott 1991;Anderson 1976), this suggests that limitations associated with pollinator behaviour are not involved, given the links between insect movements and weather (Vicens and Bosch 2009;Kilkenny and Galloway 2008;Herrera 1995;Abrol and Kapil 1986). This also suggests that the level of rainfall during flowering, which was removed as a predictor due to collinearity with temperature, is not directly implicated, as its mechanism of action would be via changes to pollinator behaviour (Lawson and Rands 2019). It is possible that precipitation during flowering could still have an effect via interaction with temperature to make fertilisation success more variable (i.e. sufficient temperatures do not always result in successful sexual reproduction if weather conditions are otherwise unfavourable). If this is the case, then there may be a longitudinal trend in variability of clonality estimates at similar latitudes, which may be something to consider in other studies of widespread clonal trees.
Other climatic variables may have a role: for instance, the action of freeze-thawing during winter induces sprouting in Fagus grandifolia and this is therefore hypothesised to increase clonal growth (Morris et al. 2014). However, this is unlikely to be important in the sampled populations, given the typical mildness of winters in the oceanic climate of the UK. In more eastern populations at similar latitudes such as those in Scandinavia or Russia, the effects of sub-zero temperatures may be involved in the promotion of vegetative propagation. The number of individuals per sample (n) and distinct multi-locus lineages therein (MLL), a proportional measure of genotypic richness (R), the complement of Simpson's index describing genotypic diversity (D*) and the number of ramets in observed MLL (NR) all provide information regarding the extent and evenness of clonality in all sites. The density of the distinct age classes, juvenile and adult, in each sample (see text for classification criteria) gives a measure of the level of recruitment and stand dominance respectively. Both the maximum distance between members of the same MLL (D max ) and the clonal dominance index (D c ) describe the spatial arrangement of clonal groups. D max illustrates the spatial extent of MLL; Dc demonstrates how aggregated individuals are relative to other genotypes (higher scores indicate more exclusive genets) The only other factor examined with a significant predictive relationship to clonal incidence was juvenile density. Quadrats with a higher number of saplings exhibited greater levels of clonality, suggesting that recent recruitment in the examined populations has been mostly vegetative in origin. This may be actual density-dependence or might simply reflect different levels of recruitment via both methods. Higher mortality rates for sexually recruited individuals compared with clones might be related to herbivory. Seedlings of T. suffer heavily from browsing by herbivores such as deer and voles (Pigott 1991). In this context, ramets may benefit from the provision of additional resources via belowground integration with the mother tree . Thus, the increase in the relative importance of vegetative propagation for recruitment may simply be the result of a failure to recruit from seed, rather than any expenditure of additional energy towards one method or another (Silvertown 2008).
It is important to note that adult density was not a predictor of clonal incidence here. As all locations had a complete canopy, adult density in this instance is likely acting as a proxy for the dominance of T. cordata within quadrats (i.e. the presence/absence of other species). Therefore, interspecific competition does not seem to affect reproductive balance in the species.
As the regression model generated here has a relatively low pseudo-R 2 value (0.66), it is clear that other unexamined factors are influencing reproductive balance. However, coppicing is not one of the factors, which is unexpected. Other species exhibit higher relative importance of vegetative propagation when under higher intensity disturbance regimes (Morris et al. 2014;Valbuena-Carabaña and Gil 2013), and T. americana has been observed directly to generate more sprouts when suffering from increased levels of herbivory (Evans and Morris 2016).
Measures of genetic diversity are in line with previous estimates (e.g. Logan et al. 2015Logan et al. , 2019, suggesting that clonality has not impacted upon this. Indeed, negative F IS in two locations shows that it has potentially increased heterozygosity due to clonal reproduction of heterozygous individuals, similarly higher levels of heterozygosity have been reported in other populations, including higher heterozygosity amongst clone genotypes compared with non-clone genotypes within the same population (Erichsen et al. 2019). The patterns of genetic diversity are not supported by the post-glacial spread from the south: the 'southern richness, northern purity' paradigm (Hewitt 2000). If a reduction in genetic diversity followed northward post-glacial expansion, then we should observe a reduction in other parameters such as polymorphism or heterozygosity, but that is not the case here.
T. cordata is more clonal than its congenerics T. platyphyllos and T. americana, although observed R appears to be similar when occurring in sympatry with the former species. Similarly, marginal populations of T. platyphyllos are also more clonal than central ones, but the degree of difference is less than in T. cordata. This suggests that similar environmental constraints, disturbance events or population histories may be affecting both species, but to a lesser degree in T. platyphyllos.
In a broader context, the proportion of unique genotypes observed here is very similar to those reported for a variety of sprouting tree species such as Populus euphratica Oliv. (Euphrates poplar), Ulmus minor Mill. (field elm) and Quercus pyrenaica Willd. (Pyrenean oak) (Valbuena-Carabaña and Gil 2017;Buiteveld et al. 2016;Schnittler and Eusemann 2010).
The character of clonal growth within populations was diverse, with many genotypes reproducing clonally and most genets occurring as small groups with restricted spatial extent. This suggests that even marginal T. cordata populations are more dynamic than might be expected, and must be recruiting from seed, albeit infrequently from a human perspective. The low frequency of members and   Occasional observations of longer distance dispersal events can be explained by opportunistic layering and windthrow producing ramets separated by relatively large distances. Individual ramets typically excluded other genets with only four quadrats (seven out of 663 individuals overall) containing any individuals that were either a product of seed or from another genet. This suggests that large clonal groups tend to out-compete other genotypes within their immediate surroundings. If any particular clonal lineage has a competitive advantage, over long timescale monoclonal stands can occur (Halkett et al. 2005;Honnay and Bossuyt 2005). This did not occur in the populations studied here. Monoclonal stands are observed in other tree species with mixed reproductive systems (e.g. Buiteveld et al. 2016;Evans and Morris 2016;Fuentes-Utrilla et al. 2014;Morris et al. 2014).
If future projections of responses to climate change in these or other taxa are to be made, then it is important to consider how these findings might influence this. The spatial arrangement of ramets and the overall number of individuals in genets have implications for the likelihood of outcrossing success, which might be important for population persistence and is relevant when considering the efficacy of selective responses (Barrett 2015;Vallejo-Marín et al. 2010). Although clonal groups were highly aggregated, their small size (in terms of both space and frequency) should not increase inbreeding and accelerate genetic diversity loss. Further evidence of this can be found in observations of pollen movements within the species, which are much higher on average than the largest clone area occurring here (Fromm 2001). In addition the diversity of clones in a population will act as a buffer to genetic diversity loss during periods of unsuitable conditions for sexual recruitment, before future climatic conditions allow for successful recruitment from seed. This of course depends on the presence of suitable conditions in all other regards, and as noted above, unexamined factors are likely involved in shifting reproductive balance.
These results suggest several avenues for further examining the role of asexual reproduction in T. cordata. While clonal recruitment disappears completely in locations where with summer temperatures allow consistent annual seed production, such as continental Europe (Logan et al. 2019;Pigott and Huntley 1981). Marginal populations are likely to retain vegetative reproduction. Habitats may however be marginal for distinct reasons; the southern and eastern range limits of T. cordata are a result of inadequate moisture (Pigott and Pigott 1993;Pigott 1991). Climatic changes in these regions are forecast to lead to extinction unless the species are able to make altitudinal range shifts (Attorre et al. 2011). A movement towards vegetative propagation under adverse conditions would impose limitations on dispersal. Therefore, it should be assessed whether reproductive balance is affected by these distinct environmental constraints (e.g. drought-related mortality rates in seedlings could exceed that of ramets).
The role of asexual reproduction in responding to disturbance should be examined more closely by assessing sprouting frequency and its relation to herbivory, controlling for the confounding influence of climate; a link has already been demonstrated in the congeneric species T. americana (Evans and Morris 2016). More broadly, this study demonstrates that natural populations provide a ing July plotted against the same response, showing that temperature during the flowering period is a significant predictor of clonality. In both instances, the line of best fit is calculated from the model's complementary log-log link function with all other covariates set to their mean values. Only significant covariates are plotted here; the full model is specified in Table 4 suitable context for detecting a relationship between incidence of clonality and its climatic constraints in trees with mixed reproductive systems. A similar approach should therefore be applied in other taxa (e.g. Valbuena-Carabaña and Gil 2017; Persson et al. 2004), given the contribution of reproductive mechanism to population persistence in a changing climate (Dodd and Douhovnikoff 2016;Sturrock et al. 2011).
In summary, this study shows that T. cordata has a variable reproductive strategy that shifts alongside typical summer temperatures. Together with previous work demonstrating a mechanistic relationship between successful pollination and ambient temperature, this demonstrates that environmental constraints are able to modify, directly or indirectly, the relative importance clonal growth versus recruitment from seed. Although other studies suggest similar links, this shows by direct comparison that canopy tree species behave in this respect as do other plant groups (e.g. Ma et al. 2013;Hautier et al. 2009). For those forest taxa with a mixed reproductive strategy, it is important to consider how this might shift because of climate change. Shifts in one direction or the other likely have major implications for long-term forecasts of responses, e.g. the likelihood of local adaptation. This demonstrates the importance of generating a better understanding of how environmental constraints interact with reproductive strategy in tree species. It also shows the value and feasibility of direct comparisons between clonal incidence and ecologically relevant variables of interest in the context of a natural experiment, using taxa with low experimental tractability, which might otherwise be ignored when studying clonality in plant species.

Acknowledgements
The authors would also like to thank all site managers and stakeholders for the permission to collect samples and invaluable advice, as well as Mark Ashton, Robert Ashton, Jordan Bilsborrow, Thom Dallimore and Andrew Devaynes for their assistance with fieldwork.
Funding This work was funded by Edge Hill University as part of an internal PhD Studentship.

Data availability
The datasets generated during and/or analysed during the current study are available from the corresponding author on reasonable request.

Declarations
Ethics approval The authors confirm that they follow the rules of good scientific practice and all ethical standards requested by the journal. Field collection was carried out in compliance with all local laws and with appropriate permissions/permits.

Conflict of interest The authors declare no competing interests.
Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article's Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http:// creat iveco mmons. org/ licen ses/ by/4. 0/.