Managing small populations—using genetic data and trial translocations to help inform suitable conservation measures for the alpine blue-sowthistle (Cicerbita alpina) in Scotland

Habitat fragmentation is causing an increasing amount of species loss around the world and creates problems at the population level. Many species are left as only small and isolated populations, which are vulnerable to genetic erosion and inbreeding depression. Here we present a study on the alpine blue-sowthistle (Cicerbita alpina). Due to intensive grazing the species is very rare in Scotland, where it occurs at only four small, montane sites, has never been reported to reproduce and is in need of conservation interventions. As the species can grow clonally it is unknown how many individuals remain and whether populations are affected by genetic isolation. We (1) quantified genetic diversity, inbreeding and between-population differentiation in Scotland and Norway using 15 microsatellite loci, and (2) experimentally translocated plants to new sites. Genetic diversity in Scotland was low (H E: 0.35; Allelic Richness: 1.84; 4 sites) compared to Norway (H E: 0.52; Allelic Richness: 2.56; 5 sites). The transplants were able to grow at new sites and are therefore not restricted to steep, montane ledges. While grazing is likely to be the main factor preventing range expansion, long-term genetic isolation has possibly further lowered population viability. To avoid local extinction of this species, conservation translocations and genetic rescue might be appropriate conservation interventions, but this needs to be further tested in a controlled environment and away from wild sites to avoid potential risks of outbreeding depression.


Introduction
Habitat fragmentation has major negative impacts on terrestrial biodiversity around the world, particularly in combination with climate change, and is threatening two in five plant species with extinction (Haddad et al 2015, IPBES 2019, Lughadha et al 2020. The direct and immediate negative impacts of fragmentation affect species' abundances and range sizes, indirect impacts can arise in the form of genetic problems following population isolation (Aguilar et al 2006, 2008, Schlaepfer et al 2018, Püttker et al 2020. The management of small and fragmented populations can be challenging and costly, as many species require continuing, species-specific interventions (Scott et al 2010). Developing effective and decisive management action may therefore be crucial to prevent declining population sizes and could ultimately avert extinction (Meek et al 2015).
Anthropogenic habitat fragmentation has a strong negative effect on plant genetic diversity (Schlaepfer et al 2018, González et al 2020. Restricted gene flow between small, fragmented populations can lead to loss of genetic variation, genetic drift, inbreeding depression, and the fixation of deleterious alleles (Leimu et al 2010, Angeloni et al 2011, Hedrick and Garcia-Dorado 2016. A large body of empirical research has demonstrated causal links between reduction in genetic diversity, elevated inbreeding, and reduced fitness in wild plant populations, characterised e.g. by lowered seed production, germination or survival rates (reviewed by Leimu et al 2006, Finger et al 2011, Pickup et al 2013. Furthermore, small populations are at a higher risk of being lost altogether due to stochastic events, such as landslides, outbreaks of pests, or grazing. To effectively manage and conserve small populations it is therefore crucial to include genetic information in conservation programmes (Gaitán-Espitia and Hobday 2021).
An example of a country that has experienced strong habitat fragmentation is Scotland. A long history of deforestation has left Scotland with only ∼4% cover of native woodlands, and intensive grazing by sheep and deer and burning regimes prevent woodlands from re-establishing (Anderson 1967, Oosthoek 2013. Consequently, many plant species in Scotland are of conservation concern (Stroh et al 2023), may require genetic restoration (Honnay and Jacquemyn 2007) and habitat conservation may simply no longer be sufficient to prevent declines due to fragmentation. Conservation translocations, the intentional movement of species for conservation purposes (IUCN/SSC 2013), might also be needed as a measure of last resort. Which conservation measures are the most appropriate and effective will depend on a good understanding of a species' genetics and ecological requirements.
We explored the genetic state of the nationally vulnerable alpine blue-sowthistle Cicerbita alpina (Asteraceae). In the UK, this species survives at only four small, inaccessible sites in the Cairngorms National Park, Scotland. While there is no documentation on this it is likely that prior to deforestation and grazing intensification, C. alpina populations were more widespread in Scotland. To guide the most appropriate conservation measures we explored whether (i) C. alpina is genetically depauperate in Scotland, (ii) remnant populations are genetically isolated, (iii) plants in Scotland can be translocated away from their natural sites.
To answer these questions, we conducted a near complete genetic inventory across the four known populations in Scotland and sampled a subset of larger and less isolated populations in Norway (which we expected to be genetically more diverse). Furthermore, we planted C. alpina into six new sites to test whether translocations can be done successfully.

Study species
Cicerbita alpina is a boreal-montane plant, which is found across Fenno-Scandinavia and in most European mountain ranges (Fries 1949, Wagenitz 1987, Meusel et al 1992. It is a perennial plant with up to 1.5 m tall panicles, flowering between July and September. The plants grow straight with each stem producing over 100 flower heads per season. Each composite flower head is about 2.5 cm wide and made up of about 40 individual violet blue flowers. The flowers are visited and presumably pollinated by bees, bumblebees, hoverflies, moths and butterflies (Marren et al 1986, Sell andMurrell 2002). Each fruit head contains up to 40 seeds (∼4 mm length × 1 mm width), which are crowned by unbranched hair and readily dispersed by wind. Seed production is between August and October.
Prior to this study, it was unknown how many individuals remain in Scottish populations. Although plants produce flowers and seeds, seedlings have rarely been recorded. It was unclear whether this was due to recruitment failure, survey errors, or other demographic reasons. The species can also reproduce asexually through clonal growth. It is therefore challenging in the field to differentiate between seedlings and new shoots of an existing plant, as it is not possible to find cotyledons in field conditions, and seedlings may consequently have been missed during surveys.

Habitat and populations 2.2.1. Scotland
In Scotland C. alpina is considered a montane species, exclusively found on wet, north-facing ledges and gullies at higher altitudes (700-1100 m a.s.l.). These provide protection from grazing animals, direct sunlight, and permanently moist but not waterlogged soils. They offer island refugia, surrounded by largely unsuitable areas of open heathlands and grasslands. However, landslides in and around some of these ledges and gullies happen frequently and the populations are at risk of being lost altogether. The species is a representative of the nationally rare tall herb community. It is possible that the species used to be more widespread in the UK prior to deforestation and heavy grazing by livestock and deer. According to Marren et al (1986), at least eight populations have been lost since 1848, with the most recent extinction occurring in Canness Glen in 1978. Historical grazing has likely led to its current restricted distribution in Scotland, current levels of grazing might hinder natural expansion, and warming temperatures could further threaten this species' existence in Scotland. Consequently, without human intervention, such as removing grazing pressure, increasing population sizes, and facilitating gene flow, long-term persistence of this species in Scotland seems unlikely.
Botanical surveys of the populations have included counting the number of flowering stems, but due to vegetative growth and plants growing in proximity it is not possible to distinguish between genetically distinct individuals. The most extensive stand at Corrie Fee (gully size ∼300 m 2 , 730 m a.s.l.) had a stem count of 239 in 1999 (Alexander 2000). Populations on Lochnagar and in Corrie Kander are dispersed over several ledges. In 1999 the number of stems recorded at Lochnagar (total ledge sizes ∼50 m 2 , 1030 m a.s.l.) was 349 stems and at Corrie Kander (total ledge sizes ∼30 m 2 , 850 m a.s.l.) 242 stems. The smallest population with a stem count of 31 is Caenlochan (one ledge of ∼4 m 2 , 850 m a.s.l.) (Alexander 2000). Recent surveys undertaken as part of this study have rediscovered a stand at Lochnagar (1120 m).

Scandinavia
Scandinavia is best suited for a comparative analysis with Scottish populations. It has the geographically closest populations to Scotland, growing at similar latitudes and altitudes. In Scandinavia C. alpina is a common and widespread plant and its National Red List status is 'Least Concern' . It is mainly found in fairly open to moderately dense birch and conifer woodlands. Shade and reliably moist but not waterlogged soils are a requirement. It can also be found in shaded places above the treeline, generally north facing. It is found from sea level to above the treeline throughout Norway, except for some Finnmark areas in the north, and throughout Sweden, except for Skåne (Solstad et al 2021). Where grazing pressure is high it is confined to cliff ledges (Bakkeveig 1983). It is climatically tolerant, ranging from very mild, wet, and almost snowless 'hyperoceanic' coasts to cold continental inland regions with seven months of snow; and with annual precipitation ranging from over 4000 mm in the most oceanic areas to less than 500 mm in parts of eastern Sweden. There is no evidence for recruitment issues in Scandinavian populations (Solstad et al 2021).

Genetic sampling
In 2015-17 we collected leaf samples for DNA extraction from all four populations in Scotland (102 samples) and five populations in Norway (52 samples), see figure 1. We attempted an exhaustive sampling in Scotland while avoiding resampling the same plant by collecting leaves across each ledge but leaving 0.5 metre between leaf samples. At Lochnagar and Corrie Kander some ledges were inaccessible, therefore we will have missed some plants there.
In Norwegian populations samples were taken across the entire population and collected at least a few metres apart to avoid sampling clones. This was to give us an indication of the levels of genetic diversity and inbreeding to expect for large, and more continuous populations. This data can then act as a baseline and indication for what we expect in a 'genetically healthy' population and be used for comparisons with Scottish populations. Populations were collected from central and southern latitudes, high and low altitudes, and large and small patches (see table 1 and figure 1). For the purpose of this study, we defined a population in Norway as a patch of plants in a certain area. The STRUCTURE analysis (figure 2) confirmed that this definition was adequate.
Due to small ledge sizes in Scotland it is more likely that we have collected clones in Scottish populations (closer proximity of collected leaf samples) compared to Norway.

Cicerbita alpina ex-situ conservation collection
Plant root stock and seeds were collected (under licence) from all four natural UK populations (2 genets from Corrie Fee, 1 Caenlochan, 2 Lochnagar, 4 Corrie Kander) and plants grown in pots and bulked up at the nursery/polytunnel of the Royal Botanic Garden Edinburgh (RBGE). Root stock from each plant can be divided repeatedly, allowing accumulation of a large number of (genetically identical) plants over time, which can be used for ecological and growing experiments.

DNA extraction, short sequence repeat (SSR) marker development, polymerase chain reaction (PCR) amplifications and genotyping
Leaf material was immediately dried and stored in silica gel. DNA was extracted from the leaves using the QIAGEN DNeasy 96 Plant Kit, following the manufacturer's protocol. All samples were screened at a total of seventeen nuclear microsatellite loci.
Enriched libraries were created by Ecogenics (Microsynth, Switzerland) with TrueSeq nano and sequenced on Illumina Miseq v2 2 × 250 bp. Microsatellites were enriched in silico using Tandem Repeats Finder, v 4.09 (Benson 1999) and primer design was done with Primer 3 (www.bioinformatics.nl/cgi-bin/ primer3plus/primer3plus.cgi/). Out of 91 microsatellite containing sequences we developed primers for 68 markers to test for polymorphism. Of these, we chose the best working markers (seventeen) which had at least three alleles. These were optimized and labelled with an M13-tag at its 5 ′ -end described by Schuelke (2000). For marker details see table 1.

Assessment of marker suitability
We tested the suitability of markers for population genetic analyses, i.e. their neutral and unlinked nature as well as null allele frequencies (Selkoe and Toonen 2006), using GENEPOP (Raymond and Rousset 1995) and CERVUS 3.0 (Kalinowski et al 2007). Significant linkage disequilibrium was observed for any pair of loci after Bonferroni correction (Hochberg 1988). All markers (except for SSR 64) had a significant excess of homozygotes. SSR 64 showed a heterozygosity of 100%, which seemed atypical and was therefore excluded from the analysis. Two markers (SSR 1 and SSR 26) had an elevated null allele frequency (0.69 and 0.75, respectively) compared to the other markers (0.11-0.55) when only analysing Norwegian samples but only SSR 26 remained higher when using the whole dataset and was therefore also excluded from further analysis, Table 1.

Repeat motif
Size range (bp) see table 1. SSR 26 was informative enough, though, to tell individuals apart and was used for that purpose. Further analyses were performed on multi-locus data from 15 microsatellites. DNA isolation has been repeated for 38% of Scottish individuals. PCR and genotyping have been repeated for all samples (Scottish and Norwegian) due to the clonal nature of the species. The genotyping error rate will therefore be minimal.

Genetic diversity estimates, inbreeding, genetic structuring, kinship, and clonal growth
Number of alleles (N A ), number of effective alleles (N E ), allele frequencies, allelic richness (A R ), observed and expected heterozygosity (H O , H E ), number of private alleles (P A ) and inbreeding coefficients (F IS ) using 10 000 permutations were calculated using SPAGeDi 1.5 (Hardy and Vekemans 2002) and GenAlEx 6.5 (Peakall and Smouse 2006).
Overall G ′ ST and Josts'D (Jost 2008) values and pairwise comparisons between populations were calculated using GenAlEx 6.5. To test for the presence of geographical groupings of related samples, we applied a Bayesian cluster analysis to all individuals using the software STRUCTURE (Pritchard et al 2000). We carried out a total of 150 runs, 1-15 clusters (K1 to K15) each repeated ten times, with the 'admixture' , 'Separate α for each Population' , and 'Correlated allele frequencies' options selected. For each run the burn-in and simulation length was 50 000 and 500 000 MCMC iterations, respectively. To identify the most likely number of distinct genetic groups K we used the ∆K statistic (Evanno et al 2005), as well as Pr[X|K] of Pritchard et al (2000). They were calculated using the online version of the software CLUMPAK (Kopelman et al 2015), http://clumpak.tau.acil). STRUCTURE output files were processed using the online version of 'pophelper' (Francis 2016), http://pophelper.com/.
To gain a better insight into the amount of selfing in populations, multilocus kinship coefficients F were calculated for all possible pairwise combinations for all plants and mean kinship coefficients for each population estimated using SPAGeDi 1.5.
Clonal growth has not been analysed systematically but when samples scored the same alleles at all 15 microsatellite markers, we assumed that these samples originated from the same genetic individual.

Trial translocations to assess plant establishment away from natural sites
To ensure that we have a good knowledge on growing and habitat requirements for this species in Scotland we conducted experimental translocations in 2017. While short term results (five year trial) will not provide evidence that the species can complete its entire life cycle at that site it helped to understand whether (i) C. alpina can grow away from its natural sites, (ii) altitude affects establishment of pregrown transplants, and (iii) plants need protection from grazing when large herbivore control is in place. We used clones originating from two plants, one from Lochnagar and one from Corrie Kander. These have been repeatedly divided resulting in large numbers of separate clones. As such we were able to repeatedly plant the same two genetic individuals into different environments. This approach allowed us to control for genetic effects impacting survival. Using only two genets (genetic individuals) has limitations and is not sufficient to provide a true representation of population level fitness. Though it needs to be stressed that due to the low number of remaining genets in populations, and the close relateness within populations (see results), one genet per population can still give an indication of the survival potential of these populations in different habitats.
Plants were controlled for pest and tested for occurrence of Phytophthora ramorum before being planted out with a small amount of soil remaining between roots. Rootball size was not measured for each plant prior to planting but is estimated to have been around 25 cm in circumference. Initially 240 rootballs were planted into three new sites: 80 at Mar Lodge, 80 at Morrone Birkwood and 80 at Corrie Sharroch, all within the Cairngorms National Park. Plants had a ratio of 80% Lochnagar to 20% Corrie Kander clones, which represents the plant material available at the time. Planting at each site was done in eight groups of five, with each group about 1-3 m apart. After grazing by voles and slugs became apparent at Mar Lodge in the first year of planting a further 75 plants were planted at this site in 2018 with additional protection by vole caging. This was done to test the effectiveness of vole cages for potential further application for larger scale conservation translocations.
All three translocation sites lie within the species' 'natural range' in Scotland, and are protected from livestock and deer either by fencing or rigorous deer stalking. The Corrie Sharroch site is situated on steep mountain ledges (750 m a.s.l.) and similar to the habitat currently occupied within Scotland. Locations at Mar Lodge (415 m a.s.l.) and in Morrone Birkwood (390 m a.s.l.) are at lower altitudes and more similar to where the species is found in Norway. Survival, rough estimate of plant size (height x width at widest point, measure for each plant), flowering, grazing pressure (estimate % of damaged leaves and stalk per plant), and plant competition (% encroaching vegetation within a 30 cm × 30 cm plot measured for each plant) were recorded over a period of four years.

Results
After removing all clones (individuals with the same genotype at 15 SSR loci), from our genetic dataset as well as samples where DNA extraction had failed, a total of 104 sampled individuals remained for further analysis, 56 from Scotland and 48 from Norway (table 2).

Genetic diversity and inbreeding
At the species level each of the 15 loci yielded between three and 19 alleles, with a total of 144. A comparison of genetic diversity over all loci and populations is given in table 2. In Scottish populations significant inbreeding (F IS ) was detected for Lochnagar and Corrie Fee, and while not significant (p < 0.1) Corrie Kander also had a lower H O than H E . In contrast, Caenlochan exhibited a homozygote deficit, possibly suggesting a recent population bottleneck, lack of selfing or an advantage for heterozygotes at this site. Values for H E ranged from 0.20 (±0.05 SE) in Corrie Fee to 0.42 (±0.06 SE) in Caenlochan. Allelic richness (A R ), based on five diploid individuals used for rarefactions, was similar across three sites but lowest in Corrie Fee 1.56 (±0.12 SE).
Comparisons with larger Norwegian populations showed that significant inbreeding is also present with similar values to Scottish populations. Genetic diversity indices (H E , A R ) though were consistently higher in Norwegian populations, on average H E was 0.56 (± 0.03 SE) and A R 3.29 (± 0.16 SE) (table 2).

Genetic differentiation
Mean pairwise genetic distances were significantly different and very high among all populations (table 3). In Scotland genetic distances were lowest between the geographically closest Corrie Kander and Caenlochan (JostD 0.35, p = 0.01; 4.6 km between populations), both located on the west side of a mountain range (The Mounth). Populations are also differentiated in Norway; even populations Theisendammen and Ferista, which are only 150 m apart, have a JostD value of 0.28 (p = 0.01). The STRUCTURE analysis identified five and nine distinct genetic clusters (K5 and K10) as the most likely solution. In both solutions this results in clustering Corrie Kander and Caenlochan into one genetic group and completely isolating Corrie Fee and Lochnagar, both of which are at least 7 km away from any other population (see figure 2). Equally, all analysed Norwegian populations form separate genetic clusters in the K = 9 solution with indication of a degree of gene flow as some individuals are not assigned to their respective population (see figure 2). The solution of nine genetic clusters remained when analysing Scottish and Norwegian population separately (data not shown).

Within population kinship coefficient
The average individual kinship coefficient within all analysed C. alpina populations is high. For Scotland Corrie Kander had the lowest kinship values, 0.42 ± 0.02 SE, and Corrie Fee had the highest average value, 0.62 ± 0.01 SE (see figure 3). Such high values can only be obtained through repeated selfing and inbreeding.
Kinship values in Norway were also high though slightly lower than Scotland, ranging from 0.37 ± 0.07 SE in Ferista to 0.23 ± 0.01 SE in Stugudal (see figure 3).

Trial translocations 3.5.1. Survival and plant size
Of 315 planted C. alpina clones, 31% survived four years after planting, (44% in Mar Lodge, 30% in Corrie Sharroch, and 6% in Morrone Birkwood, table 4). Poor performance at Morrone Birkwood is likely due to poor micro-siting (e.g. planting in waterlogged soils). Plant survival stayed similar between years 3 and 4 at        SE 195 and 1198 cm ± SE 197, respectively) compared to Morrone Birkwood (208 cm ± SE 71). At Corrie Sharroch this was particularly true for plants located in a gully with steep walls, which may have sheltered the plants from wind, direct sunshine and dehydration. Third year size data for Corrie Sharroch was not available as most plants were heavily grazed down to the stem. Plant size at all sites increased over the years which may indicate that those plants surviving are starting to establish.

Flowering
Overall

Are small populations genetically depauperate and isolated?
We have identified 56 genets in Scotland, although we are likely to have missed some, it is evident that fewer than 100 C. alpina genets remain in Scotland, which is a very low number. Moreover, genetic diversity is relatively low and kinship high within populations compared to larger Norwegian populations. Another study on C. alpina shows that Scandinavia has the highest AFLP diversity, and that other European regions are less diverse in comparison (though excluding Scotland) (Michl et al 2010). Extended periods of selfing, very low numbers and resulting genetic drift could explain the observed high kinship and low genetic diversity within populations in Scotland. High inbreeding and kinship values are also characteristic for Norwegian populations, suggesting that selfing is not unusual for the species. Self-compatible clonal plants will have increased levels of geitonogamy (pollination between different flowers of the same genetic individual) as even high visitation rates by pollinators will result in high selfing rates if clones are large or where clonal diversity of (small) populations is low.
Strong genetic differentiation between populations indicates that gene dispersal rates in C. alpina may be lower than expected for a wind-dispersed and insect-pollinated species (Dick et al 2003, Lander et al 2010, Finger et al 2014. Population Corrie Fee has a slightly different leaf shape compared to other populations, see figure 4. Though leaf shape has not been systematically analysed it may be a further indication that this populations has been isolated and restricted to within population reproduction since many generations. Limited gene flow was also detected between populations in Norway, even when only 150 m apart. Occasional, long-distance gene flow events there may though be enough to prevent genetic isolation in more continuous habitats. Cicerbita alpina can reproduce clonally which further affects its ability to persist and expand in Scotland, both positively and negatively. While clonality increases population survival, predominant clonal reproduction can affect long-term phenotypic evolution which could impact on the species' abilities to adapt to changing environments (Orive et al 2017). It has been suggested in previous research that increased clonal growth might be a response to small population size (Duwe et al 2017). The smallest population Caenlochan though had the lowest amount of clonal growth (eleven of the 13 collected leaf samples belonged to different genets). Though one reason may simply be the lack of space on a 4 m 2 ledge. Where ledges are larger (Corrie Kander and Corrie Fee), clonal growth is more frequent. Therefore a mechanistic link between reduced population size and increased clonality is unlikely and for C. alpina this rather reflects that conditions for successful germination and seedling establishment are rare.
The observed high inbreeding values, strong between population structure in both Scotland and Norway suggest that the species is naturally selfing. It is therefore possible that through many generations of selfing it has purged its genetic load and that inbreeding does not necessarily lead to inbreeding depression (Crnokrak and Barrett 2003, Hedrick and Garcia-Dorado 2016, Glémin et al 2019. Whether inbreeding depression might be an issue for this species needs to be further investigated.

Can plants establish away from wild sites?
Our trial translocations demonstrated that transplant survived at high (750 m a.s.l.) and low altitudes (400 m a.s.l.), but that grazing can be an issue for plant establishment and survival, even at sites where grazing by large herbivores has been almost completely eliminated. A study analysing 249 plant species translocations worldwide showed that on average survival for translocated plants was low with 52%, flowering 19% and fruiting 16% (Godefroid et al 2011). In this study, success rates for C. alpina were even lower, with only 31% survival and 13% of surviving plants flowering. This may in part be due to the low genetic diversity of plants (only two genetic clones used). Grazing at lower altitudes had a strong negative impact on plants from the first year of planting. While the higher altitude site (Corrie Sharroch) was only mildly affected by grazing in the first two years and year four, it did get heavily damaged in its third year. Thus, our study only partially confirms the notion that montane sites provide refuges from herbivory (Scheidel and Bruelheide 2001). In the absence of changing grazing levels, future translocations will likely rely on additional protection through (expensive and often inefficient) fencing to limit grazing pressure (Scheidel andBruelheide 2001, Fenu et al 2016).
Current distribution patterns in Scandinavia demonstrate that (under lower grazing pressure) C. alpina grows at lower altitudes and warmer temperatures. We therefore expected that grazing was the main driver for the species' retreat onto high elevation ledges (Scheidel andBruelheide 2001, Oosthoek 2013). It is likely though that changing plant species compositions (through changed land management) has further contributed to this retreat (Alexander et al 2015). While low plant competition appeared to be linked to higher survival in our translocation experiment, this contrasts partly with Norway, where C. alpina creates patches able to compete and persist within luxuriant, species-rich assemblages. Whether this increased ability to compete is due to higher genetic diversity, more suitable environments, or due to other reasons is currently unclear.
It should be noted that our observations are based on four years of monitoring which is too short to indicate whether plants are likely to persist long-term in these sites and able to produce seedlings (Godefroid et al 2011, Albrecht et al 2019, Bubac et al 2019. Varying plant performance between years and sites as shown in our data, further emphasise the need for long-term datasets. Furthermore, using two single genetic individuals for these trial translocation might not be sufficient to provide a true representation of population level fitness. The trials have demonstrated that plants can grow at new sites, in habitats more typical of where Norwegian C. alpina are found which will significantly increase site options for future conservation translocations.

Conservation recommendations
Increasing population sizes and creating or maintaining gene flow between populations can halt fitness declines and the extinction of small populations (Ellstrand 2014, Ellstrand andRieseberg 2016). Limiting grazing around wild populations might help plants to expand but as the surrounding vegetation has become largely unsuitable (except possibly on other small ledges) it is questionable whether deer control by itself will allow establishment and/or expansion into surrounding areas. Given that in some cases habitat conservation may simply no longer be sufficient to prevent declines, as is the case for C. alpina, there is an increasing interest in the potential of conservation translocations, the intentional movement of species for conservation purposes (IUCN/SSC 2013). Such translocations should only be considered if suitable habitat is available. Our results have shown that a small proportion of pre-grown transplants can grow at a range of new sites when micro-siting is correct and conservation translocations are therefore likely to succeed. This needs to be further tested at a larger scale though. Such experiment would ideally also have to determine whether transplants can complete the full life cycle, produce seeds which can then germinate, establish and grow up to form new plants and form a viable population, though this would likely take decades. Other trials could also include growing new populations from seed (if enough can be generated), mimicking the natural selection process during germination, establishment and first growth as much as possible. This approach would allow to immediately determine whether the new habitat is suitable for the completion of the entire life cycle, and not only for the survival of planted adults.
To increase genetic diversity and lower inbreeding it would also be beneficial to mix plants from different populations and plant them into new sites where they have enough space to expand. We recommend doing controlled crossing experiments (using Scottish population as well as Norwegian and European provenance) to determine whether mixing between populations is likely effective for this species before planting them out. Such a conservation intervention would resemble genetic rescue, the supplementation of genetically impoverished populations with new individuals (or genotypes) with the purpose of reducing genetic load (Thrall et al 1998, Bell et al 2019 and enhancing population viability (Tallmon et al 2004). Planting into new sites rather than introducing genes into existing populations would further avoid any potential risks of outbreeding depression by undermining evolved local adaptation among recipient populations (Edmands 1999, Tallmon et al 2004, Barmentlo et al 2018 and genetic swamping (Hufford and Mazer 2003). These sites will need to have significant deer control in place and be part of wider habitat restoration programmes to create suitable habitat for the species long term and a more balanced herbivore to plant balance.

Data availability statement
All data that support the findings of this study are included within the article (and any supplementary information files).