See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/369592001

Habitat fragmentation affects climate adaptation in a forest herb, now

published @https://besjournals.onlinelibrary.wiley.com/doi/abs/10.1111/1365-
2745.14225

Preprint · March 2023

DOI: 10.48550/arXiv.2303.15712

CITATIONS READS

0 249

4 authors:

Frederik Van Daele Olivier Honnay

Ghent University KU Leuven
13 PUBLICATIONS 26 CITATIONS 449 PUBLICATIONS 19,299 CITATIONS

SEE PROFILE SEE PROFILE

Steven B Janssens Hanne De Kort

Meise Botanic Garden KU Leuven
340 PUBLICATIONS 4,197 CITATIONS 39 PUBLICATIONS 1,133 CITATIONS

SEE PROFILE SEE PROFILE

All content following this page was uploaded by Frederik Van Daele on 18 May 2023.

The user has requested enhancement of the downloaded file.

Habitat fragmentation affects climate
adaptation in a forest herb

Local adaptation in Primula elatior

Frederik Van Daele1, Olivier Honnay1, Steven Janssens 2,3, Hanne De Kort1

1 Department of Biology, KU Leuven, Leuven, Belgium

2 Meise Botanic Garden, Nieuwelaan 38, BE-Meise, Belgium

3 Leuven Plant Institute, KU Leuven, Leuven, Belgium

Correspondence:

Frederik Van Daele

[email protected]

1
Abstract
1. Climate change and the resulting increased drought frequencies pose considerable threats to
forest herb populations, particularly where additional environmental challenges jeopardize
responses to selection. Specifically, habitat fragmentation may impede climate adaptation through
its impact on the distribution of adaptive genetic variation, and cause evolutionary shifts in mating
systems.
2. To assess how habitat fragmentation disrupts climate adaptation, we conducted a common
garden experiment with Primula elatior offspring originating from 24 populations sampled along a
latitudinal gradient with varying climate and landscape characteristics. We then quantified a range
of vegetative, regulatory, and reproductive traits under distinct soil moisture regimes to evaluate
imprints of local adaptation and phenotypic plasticity. Additionally, we conducted a more extensive
field campaign in 60 populations along the same latitudinal gradient to evaluate the potential
evolutionary breakdown of reciprocal herkogamy.
3. For large, connected populations, our results demonstrated an evolutionary shift from a drought
avoidance strategy in southern populations to a drought tolerance strategy in northern
populations. However, habitat fragmentation disrupted climate clines and the adaptive responses
to drought stress in key traits related to growth, biomass allocation and water regulation.
Additionally, our findings indicate the onset of evolutionary breakdown in reciprocal herkogamy
and divergence in other key flower traits. The disruption of climate clines, drought responses, and
adaptations in mating systems contributed to a substantially diminished flowering investment
across the distribution range, with the most pronounced effects observed in southern fragmented
populations.
4. Synthesis: We present novel empirical evidence of how habitat fragmentation disrupts climate
adaptation and drought tolerance in a wide range of traits along the range of the forest herb
Primula elatior. These findings emphasize the need to account for habitat fragmentation while
designing effective conservation strategies in order to preserve and restore resilient meta-
populations of forest herbs amidst ongoing global changes.

strategies, such as assisted translocation and

1. Introduction targeted gene flow.
Rapid climate change is increasingly disrupting
long term imprints of natural selection on plant Climate clines play a significant role in shaping
physiology and phenology (Becklin et al., 2016; local adaptation in plant species. As temperatures,
Cleland et al., 2007). As climate change outpaces precipitation, and other climatic variables differ
the evolutionary rate of local adaptation, population across various geographic regions, plant
viability becomes increasingly compromised (Jump populations have evolved to adapt to the local
& Peñuelas, 2005). Habitat fragmentation, a result environmental conditions they reside in (Franks et
of rapid land-use changes, can exacerbate the al., 2014). These climatic adaptations can manifest
impacts of climate change by constraining the as changes in genetics, physiology, phenology, or
ability of species to migrate and adapt to the as phenotypic plasticity. For example, forest herbs
changing environment (Honnay et al., 2002; Skov evolved distinct strategies for reproduction (Graae
& Svenning, 2004). Furthermore, the independent et al., 2009), growth (De Frenne et al., 2011),
and interactive effects of climate drivers and phenology (Bucher et al., 2018; Cleland et al.,
habitat fragmentation on local adaptations may 2007), resource allocation (Qi et al., 2019;
have significant implications for plant viability and Rosbakh et al., 2015), or investment in specialized
fecundity. Disentangling the effects of these global leaf morphological or stomatal adaptations (C. Liu
change drivers on local adaptations is essential in et al., 2018; W. Liu et al., 2020), driven by factors
order to inform conservation and management such as the annual growing degree days, cold
tolerance, heat stress, and water availability (Skov

2
& Svenning, 2004). Phenotypic plasticity can also 2020; Zellweger et al., 2020) and could disrupt the
be relevant for the establishment and persistence phenology of spring blooming forest herbs (Kudo et
of populations, especially in the face of increasing al., 2008). Furthermore, non-climatic selection
climate extremes and drought frequencies (De Kort pressures in fragmented habitats could also disrupt
et al., 2020). However, adaptive divergence climate adaptation due to pleiotropic effects
between populations is often mediated by (Armbruster, 2002; Etterson & Shaw, 2001;
counteracting factors such as gene flow, genetic Strauss & Whittall, 2006). Since both habitat
drift, biogeographical history, community fragmentation and climate drivers can influence the
dynamics, and interacting environmental evolution of plant traits, it is thus essential to
pressures (Blanquart et al., 2013; Kawecki & Ebert, disentangle how they interactively shape local
2004). adaptation and to determine to what extend habitat
fragmentation disrupts the adaptive potential of
Habitat fragmentation can lead to increased plant populations.
isolation, demographic stochasticity, and altered
environmental and biotic interactions, all of which Larger distances between remaining habitat
can assert distinct selection effects and impact the patches and an unsuitable landscape matrix
evolutionary dynamics of plant populations (Dubois reduces the dispersal potential of seed propagules
& Cheptou, 2017). The adaptive capacity of and decreased the potential for pollination,
species in response to climate change is thus asserting evolutionary pressures on both dispersal
influenced by complex eco-evolutionary feedback and mating mechanisms (Cheptou et al., 2017;
loops involving interactions between various global Opedal, 2019). A hostile landscape matrix
change drivers (Hanski et al., 2011). These increases dispersal risks and may result in
feedback effects can be antagonistic, synergistic, selection pressures that prioritize reproduction
or additive (Coors & DeMeester, 2009), resulting in within habitat fragments while decreasing seed
unpredictable adaptive responses to climate dispersal investment (Bonte et al., 2012). On the
change across a species' native range (Jump & one hand, limitation of the seed dispersal potential
Peñuelas, 2005). For example, the reduced can lead to selection pressures which improve
genetic variation and increased inbreeding often dispersal mechanisms in order to avoid inbreeding
observed in fragmented populations may lead to a depression (Cote et al., 2017; Olivieri et al., 1995).
reduced adaptive potential of plant populations Smaller plant population sizes, lower flower
(antagonistic effect), consequently jeopardizing densities and increased spatial isolation disrupt
adaptive stress responses (Leimu et al., 2010). In plant–pollinator mutualisms in fragmented habitats
contrast, a lower gene flow can, in some cases, (Aguilar et al., 2006; Eriksson & Ehrlen, 2001;
accelerate local adaptations by preventing the Gómez-Martínez et al., 2020; Kwak et al., 1998).
influx of mal-adapted genes or swamping of locally Altered or impoverished pollinator communities
adapted genes, resulting in synergistic eco- could lead to selection pressures for traits that
evolutionary effects (Savolainen, Lascoux, and improve pollinator attraction or co-evolutionary
Merilä, 2013). Habitat fragmentation could also flower syndromes (Gumbert, 2000). For example,
lead to increased variability in abiotic and biotic heritable local adaptations of “bull’s-eye” patterns
conditions, which could select for strategies that have been documented, in which the petal bases
enable species to cope with heterogeneity absorb UV light while the petal apices reflect it,
(Cheptou et al., 2017). Furthermore, fragmented effectively directing pollinators towards the
habitats are often characterized by distinct land- reproductive structures (Koski & Ashman, 2013).
use histories, are subject to increased edge Moreover, disruption of plant-pollinator networks in
effects, which can lead to distinct microclimates, fragmented habitats can lead to increased
irradiation regimes, increased soil nutrient selection pressures on herkogamy, the relative
contents, and modified interspecific interactions positioning of stigmas and anthers within flowers
(Baeten et al., 2015; Xiao et al., 2016). These (Jacquemyn et al., 2012; Mitchell et al., 2008).
altered microclimates and interspecific conditions Research indicates that noticeable evolutionary
in fragmented habitats could impose selection changes in herkogamy can rapidly emerge in
pressures on growth patterns (Blondeel et al., response to severe disruptions in plant-pollinator

3
networks (Opedal, 2019; Opedal et al., 2017). rivers (hydrochory) in riparian habitats occur as
Nevertheless, the pace of this evolutionary process well (Endels et al., 2004; Vittoz & Engler, 2007). P.
is expected to be more gradual in long-lived elatior generally requires seed dormancy to ensure
species that heavily rely on clonal reproduction (De that seeds only germinate under favourable
Witte & Stöcklin, 2010). The evolutionary conditions (Baskin & Baskin 2004; Taylor et al.
breakdown of herkogamy can lead to feedback 2008; Baskin 2014). This process typically involves
loops, as reduced herkogamy can drive increasing a period of exposure to low temperatures, which is
levels of autonomous selfing (Herlihy & Eckert, further modulated by changes in irradiation levels.
2007; W. Zhang et al., 2021). This may be Furthermore P. elatior has found to be especially
beneficial when pollination is limited, but it can also sensitive to soil waterlogging and drought (Taylor
increase inbreeding and reduce genetic diversity et al., 2008; Whale, 1983). P. elatior has distylous
(Jacquemyn et al., 2012; Porcher & Lande, 2005). flowers with reciprocal herkogamy and pollination
is mainly performed by bumblebees (Dickinson,
Given the profound yet complex ways in which both 1990).
habitat fragmentation and climate change can
influence plant evolution, and considering the To evaluate the interactive effects of climate and
substantial global variation in habitat fragmentation habitat fragmentation on the mating system and
levels, we designed our research study to adaptive trait evolution, we collected seeds for a
disentangle their individual and interactive effects common garden experiment of P. elatior along a
on the adaptive potential of plant populations. paired latitudinal gradient (Fig. 1; De Frenne et al.
Specifically, we conducted a common garden 2013) involving one highly fragmented and one
experiment complemented by an extensive field well-connected forest patch (large population in
survey to elucidate the impact of habitat contiguous habitat), ranging from southern France
fragmentation on the evolutionary trajectories of up to northern Denmark. The paired sampling
natural Primula elatior populations. Our study design was chosen to disentangle landscape
aimed to: (objective 1) quantify the phenotypic fragmentation from climate effects along the
variation driven by climate adaptation across a distribution range. Seeds were collected from 15
north-south cline, including responses to periodic individuals in all 24 populations (12 pairs)
droughts; (objective 2) assess the extent to which originating from respectively southern (southern
habitat fragmentation disrupts climate adaptation, France), central (northern France up to the North
particularly under drought stress; and (objective 3) of Belgium) and northern (Denmark) Europe (four
evaluate the effects of habitat fragmentation on pairs in each region). The number of individuals in
evolutionary shifts in the mating system. each population were counted as a proxy for
population size, and seeds were collected from
2. Methods widely spaced individuals. Herkogamy was
determined in 60 populations along the paired
2.1. Study system and sampling latitudinal gradient with a total of 2116 evaluated
Primula elatior subsp. elatior (Primulaceae family) individuals and three flowers per plant (n = 6896)
is a European forest herb with a native distribution in order to quantify the potential evolutionary
range from southern France up to Northern breakdown of self-incompatibility.
Denmark (Taylor et al., 2008; Van Daele et al.,
2021). Primula elatior is dispersal limited, self- 2.2. Common garden
incompatible and is regarded as a representative A total of ~30 seeds originating from at least 15
model species for the mainly slowly colonizing parents (F1) from the 24 populations were spread
broadleaf forest herb layer (Van Daele et al., 2021; evenly in petri dishes filled with perlite and water
Verheyen et al., 2003). Rhizome-driven vegetative for two treatments (i.e. with and without
colonization is very limited (Baeten et al., 2015) vernalisation), resulting in a total of 976 petri
and seed dispersal is mainly determined by short dishes. The effect of vernalisation was evaluated to
range mechanisms such as gravity (barochory) monitor the effects of increasing minimum
and wind (rolling anemochory). Occasional seed temperatures during winter times on germination
herbivory (endozoochory) or transportation by success and germination lag time. In the

4
vernalisation treatment, seeds were cooled at a due to its correlation to the canopy phenology
maximum temperature of 6° C for 7 weeks. (Baeten et al., 2015; Richardson et al., 2006).
Germination was enabled under a controlled
climate regime of 12 hours light at 21° C and 12 To determine evolutionary adaptation drivers on
hours dark at 16° C (C. C. Baskin, 2014). plant persistence, we quantified the adult growth
rate and resource allocation traits such as the root-
Seedlings were randomized in a controlled
common garden greenhouse experiment with four
offspring (F2) originating from each of 15 parents
(F1) for each population (24) in two treatments
(2880 individuals) around ~ the 28th of February in
2020. In a first treatment (control), the water
availability neared field capacity while staying
below allowable depletion to avoid drought stress
(Appendix S1). Two drought periods were induced
near the pre-determined wilting point (18.4%± SE
4.87% soil water content) in the second treatment
to evaluate responses to drought stress, one
during the autumn of 2020 (~25% soil water
content) and one during the spring of 2021
(~37.5% soil water content).

2.3. Response variables Figure 1: Data collection locations of Primula elatior flowers and
To evaluate evolutionary adaptation drivers related seeds. Herkogamy data was collected in all 60 locations. The seeds
used for the common garden experiment originated from 24 of
to plant establishment and phenology, we these locations.
evaluated the germination lag time, the
shoot ratio and specific leaf area (SLA). The adult
germination success, juvenile growth rate, and the
plant height was reassessed the 15th of June 2021
flowering phenology. The germination lag time was
to evaluate the adult growth rate (~337 days) while
evaluated every three days and the germination
resource allocation traits were determined at the
percentage was determined after 6 weeks (no new
end of the experiment (august 2021). Shoots were
germination for a week). Following the germination
separated at the growing point and belowground
experiment, seedlings were measured at the start
biomass was rinsed with water to remove soil.
of the common garden experiment (28th of
Biomass was oven-dried for 48 hours at 70°C and
February 2020). The length of seedlings was re-
the weight was determined with a precision of 10-2
evaluated at the 21st of April in 2020 (after ~53
g. The specific leaf area (SLA; leaf area/leaf mass)
days)in order to determine the seedling growth rate
was determined at the same time. First, the leaf
(mm/day). Juvenile growth was reassessed the
area was calculated based on leaf scan images
13th of July in 2020 when plants reached adulthood
with the R package “Leaf Area v. 0.1.8” (Katabuchi
(~83 days). Missing values of the seedling growth
& Masatoshi, 2015) and leaves were subsequently
rate, juvenile growth rate, and overall juvenile
oven dried for 48 hours at 70°C and weighted with
growth (from seedling to adult) were assigned the
a precision of 10-3 g.
population mean. As each of these variables had
very similar growth rate patterns, they were To evaluate micro-morphological leaf structure
reduced in dimensionality with principal component adaptations, we evaluated both abaxial and adaxial
analysis (PCA) with varimax rotation. The first PC stomatal density, as well as the abaxial glandular
axis explained 78.74% of the variance with a sum trichome density. First, abaxial and adaxial leaf
of squared loadings of 2.36 and the three metrics imprints were made with clear nail polish. For each
had a high correlation with the first PC axis (r > 0.7). imprint, three pictures were made with a Keyence
Flowering phenology or the flowering lag time was VHX-5000 microscope magnified by 300x with a
determined based on the cumulative degree days 688 x 516 μm surface area (0.36 mm²) for the
warmer than 4°C (CDD4) starting from 1 January adaxial side and 500x with a 1145 x 859 μm

5
surface area (0.98 mm²) for the abaxial side. To 2.7.3” (Ooms, 2021), “imager v. 0.42.13”
classify and count the abaxial and adaxial stomata, (Barthelme, 2022), “exifr v. 0.3.2” (Dunnington &
and the abaxial glandular trichomes, custom Harvey, 2021), “hexView v. 0.3-4” (Murrell, 2019),
convolutional neural network models were trained “raster v. 3.4-13” (Hijmans, 2019),
(Appendix S2), optimized and evaluated using landscapemetrics v. 1.5.4" (Hesselbarth et al.,
“python v. 3.9.7” on the Flemish supercomputer 2019), SpaDES (Chubaty & McIntire, 2022),
network (Flanders Research Foundation, 2022) following the false colour bee view reconstruction
with the Tensorflow 2.8.0 (Abadi et al., 2015) and recommendations of Lunau, Verhoeven and Ren
Keras 2.8.0 (Chollet & others, 2015) libraries, (2018). The median flower colour value (0-255) of
based on the guidelines outlined in the “from leaf UV red, green and blue bands were reduced in
to label workflow” (Meeus et al., 2020). To evaluate dimensionality using a PCA with varimax rotation.
the neural network model quality, true positives The first PC axis explained 52.97% of the variance
(TP), false positives (FP) and false negatives (FN) with a sum of squared loadings of 1.59. The first
were manually determined for a random stratified PC axis was 51.12% correlated to UV red
selection of 300 classified images for each model. variability, while the blue band was 86.1%
Overall model performance was evaluated by the correlated and the green band was 76.59%
Jaccard index (i.e. threat score) of the binary correlated. Stigma height (SH) and the anther
confusion matrix:
𝑇𝑇𝑇𝑇
. The threat scores of height (AH) were separately determined in the field
𝑇𝑇𝑇𝑇 + 𝐹𝐹𝐹𝐹 + 𝐹𝐹𝐹𝐹
for 60 populations along the range (see study
the CNN image classification models were 0.92 ±
system and sampling) with a 0.01 mm precision
0.07 SD, 0.73 ± 0.20 SD, and 0.81 ± 0.18 SD
digital caliper to evaluate herkogamy based on the
respectively for the abaxial stomata, adaxial
guidelines of Brys and Jacquemyn (2015). The
stomata, and the abaxial glandular trichomes,
potential breakdown of biochemical self-
indicating a high degree of similarity between the
incompatibility systems (Bawa et al., 2011;
predicted and observed counts.
Dickinson, 1990) was evaluated with manual self-
To evaluate the evolutionary adaptation drivers on pollination to circumvent potential barriers caused
mating mechanisms, which in turn determine the by heteromorphic self-incompatibility. To this end,
reproductive capacity, we determined the flower three flowers per plant (all plants in the control)
investment, bee view, herkogamy, and self- were separately enclosed in string sealed tea bags
compatibility. Flower investment was determined prior to flower opening. During the reproductive
based on the total count of produced flowers, the stage, flowers were carefully self-pollinated with
flower accessibility as determined by the mean disposable cotton swab and re-enclosed in the
flower opening duration (of three stalks), the total sealed tea bags. Finally, when seed pods
count of produced stalks, and the average stalk contained at least 1 developed seed they were
height. Missing values of these flower investment scored as self-compatible (binary scoring of each
metrics were imputed with the population mean seed pod).
and reduced in dimensionality with PCA with
varimax rotation and these variables were highly 2.4. Explanatory variables
correlated with the first PC axis. To assess the local To determine climate variability between
adaptation of visual pollinator signalling, we populations, 19 bioclimatic variables (WorldClim
evaluated the three main bandwidths experienced version 2.1; Fick and Hijmans, 2017) were
by the main pollinators of P. elatior, namely bees extracted in relation to the coordinates of the 60
from the genus Bombus (Schou, 1983). A macro- habitat locations with R package “raster v.3.4-13”
photography studio was constructed and natural (Hijmans, 2019). A PCA with varimax rotation was
light irradiation was standardised with a NISSIN used to reduce dimensionality and prevent
MF-18 ring flash for the RGB bandwidth images, multicollinearity with the R package “psych v.2.2.5”
while a Convoy C8 365nm UVA with ZWB2 UV/VIS (Revelle, 2022). Based on the acceleration factor,
filter was used for UV bandwidth images. A custom which was determined with the R package
algorithm was written for image segmentation, “nFactors v.2.4.1” (Raiche & Magis, 2020), the first
band separation, and false colour image PC was retained as an explanatory variable. The
construction using the R packages “magick v. first PC of the climate variables explained 64.21%

6
of the variance and had a sum of squared loading of floral traits to better decipher the distinct
of 12.20. The first PC displayed a strongly positive evolutionary selection pressures on mating
correlation with temperature and precipitation mechanisms associated with reciprocal
metrics and a strong negative correlation with herkogamy.
temperature seasonality (Appendix S2).
2.5. Modelling
Habitat fragmentation was evaluated based on a To evaluate climate driven phenotypic evolution,
set of ecologically representable landscape metrics including drought response adaptation (O1), and to
of broadleaf forest (Copernicus Land Monitoring quantify the disrupting evolutionary influences of
Service, 2018) with a grid size resolution of 100m habitat fragmentation on climate adaptation (O2),
and the surveyed population size. The patch size we modelled the three-way interaction of the first
(area metric), distance to edge (edge metric), and climate PC, the first PC of habitat fragmentation
perimeter-area ratio (shape metric) were metrics, and the drought treatment or vernalisation
determined on a habitat patch level with the R factor (for germination traits) as drivers of adaptive
package “raster v.3.4- 13” (Hijmans, 2019) and trait divergence. Mixed models were constructed
“landscapemetrics v.1.5.4” (Hesselbarth et al., with in “lme4 v. 1.1-27.1” and model assumptions
2019). To determine broadleaf forest class metrics, were evaluated and adjusted accordingly
buffer radius distances of 2000m, beyond which (Appendix S3) with the R package “performance v.
little or no pollen and seed dispersal is expected 0.9.2” (Lüdecke et al., 2021) and “DHARMa v.
were calculated from the patch centroid 0.4.6” (Hartig, 2022).
(Jacquemyn et al., 2002; Van Rossum, 2008).
Within these landscapes, we calculated the mean The germination lag time was modelled with a
patch area (area metric), aggregation index negative binomial generalized linear model and log
(aggregation metric; He, DeZonia and Mladenoff, link function to control overdispersion of the
2001), cohesion (aggregation metric; Schumaker, deviating residual distribution with the R package
1996), division (aggregation metric; Jaeger, 2000), “lme4 v. 1.1-27.1” (Bates et al., 2015). The
mean gyrate (area and edge metric; Keitt, Urban germination success (rate) was modelled with a
and Milne, 1997), number of patches (aggregation quasibinomial generalized linear model with a logit
metrics), and the class percentage of the link to control for overdispersion, weighted by the
landscape (area and edge metric). These 11 total used seeds for each petri dish. The juvenile
metrics were reduced in dimensionality with PCA growth (PC1), the mean flowering phenology
and a varimax rotation. Based on the acceleration (CDD4; mean of three stalks), and adult growth
factor, the first PC was retained as an explanatory rate were modelled with linear mixed-effects
variable. All metrics had a correlation coefficient (r) models with the parent ID (F1) as random factor.
of at least abs(0.5) and the first PC explained 55% Models had a normal distribution of the residuals
of the variance with a sum of squared loadings of and random effect, homogeneity of the residual
6.07 (Appendix S2). Lower values of this variance (homoscedasticity) was observed, and
component represent higher levels of habitat the model-predicted data was in line with the
fragmentation (decreased populations size, patch observed data. The root-shoot ratio (g) and SLA
sizes, distance to edge, aggregation, cohesion, (mm2mg-1) was modelled with linear mixed-effects
gyrate, and landscape proportion; and an models with the parent ID (F1) as random factor.
increased perimeter-area ratio, division, and The root-shoot ratio was square root transformed
number of patches). and the SLA was log transformed to obtain a
normal distribution of the residuals. The abaxial
Vernalisation and drought treatments were and adaxial stomatal density, and abaxial glandular
indicated as binary factors to explain respectively trichome density was determined with linear mixed-
the phenotypic plasticity of vernalisation on effect models with the plant ID (F2) and parent ID
germination traits and phenotypic plasticity under (F1 origin) as nested random factors. For each
drought stress of relevant common garden traits. model, the response was square root transformed
Long-styled (L) and short-styled (S) flower morphs to obtain a normal distribution of the residuals. The
were incorporated as binary factors in the analysis climate partition (phenotypic trait variation), partial

7
effect of drought stress (phenotypic plasticity), and effects. The response to drought stress was
the contribution of their interaction (variability of replaced by vernalisation for the standardised
phenotypic plasticity along the range) were variance partitioning of germination, while it was
evaluated in order to answer the first hypothesis replaced by the morph type for the bee view,
(O1). Evolutionary processes (additive, synergistic, herkogamy, and biochemical self-incompatibility.
or antagonistic) driven by habitat fragmentation Marginal effects of individual model terms were
that disrupt climate adaptation (O2), were graphically represented using “sjPlot v. 2.8.9”
evaluated based on the contribution of the (Lüdecke, 2021) and “ggplot2 v. 3.3.6” (Wickham,
interaction effect of climate (PC1) and habitat 2016) with habitat fragmentation as moderator
fragmentation (PC1), and additionally by the variable using all observed values (Aiken & West,
contribution of the interaction effect of habitat 1991; Dawson & Richter, 2006). Contributed
fragmentation (PC1) and drought stress. proportions of the explained variance (Lüdecke et
al., 2021; D. Zhang, 2021) of separate variables
The extent to which habitat fragmentation drives were based on their proportion relative to the three-
evolutionary shifts in the mating system (O3) as way interaction effect. The contribution of two-way
determined by the flower investment was interactions were determined as contribution
evaluated based on the partial habitat proportions relative to the three-way interaction
fragmentation contribution. The shifts in mating effect minus the contribution proportion of both
systems as determined by the bee view, separate variables. Finally, the contribution of
herkogamy, and self-pollination was evaluated single effects and two-way interaction effects was
based on the habitat fragmentation partition, the standardised to eliminate potential deviations of
morph type partition, and the interaction effect the total contribution due to remnant confounding
partition. To evaluate how habitat fragmentation effects such as multicollinearity and suppression
drives evolutionary shifts in flowering investment, (Lai et al., 2022). The standardised variance
we first assessed the three-way interaction partitions and statistical significance of separate
between climate PC1, habitat fragmentation PC1 terms and interaction terms were reported in venn
and drought stress (PC1) with a linear mixed- diagrams using “VennDiagram v. 1.6.20” (Chen,
effects model with the parent ID (F1) as random 2018). Independent or interactive fixed effects on
factor. The evolutionary shifts in bee view phenotypic trait variation were considered in detail
(pollinator attraction mechanism), herkogamy, and when the effect was significant (p>0.05) and had a
biochemical self-incompatibility was evaluated with considerable contribution to the overall explained
the three-way interaction effect of the climate PC1, fixed effect variance (>5%), as determined by the
habitat fragmentation PC1, and the flower morph variance partitioning.
type. The three-way interaction effect on the mean
bee view (mean of three flowers) was evaluated We utilized redundancy analysis to examine the
with a linear mixed-effects model with the parent ID unique and shared influences of habitat
(F1) as random factor. The mean herkogamy per fragmentation and climate on multivariate
plant was modelled with a generalized linear model phenotypic trait variation and phenotypic plasticity
and a gaussian error distribution with log link. (drought responses) using the vegan package
Biochemical self-incompatibility was evaluated version 2.5-7(Oksanen et al., 2013). First, values
with a weighted binomial generalized linear mixed- were averaged on the parent ID level (F1) and
effects model with logit link, the plant ID (F2) as treatments were spread into a wide format to
random factor, and evaluated flower count as create separate phenotypic plasticity variables
weights. Convergence failure was here addressed using “tidyr v. 1.1.3” (Wickham, 2021) while
with the allFit optimizer algorithm in “lme4 v. 1.1- missing values were imputed with the population
27.1” (Bates et al., 2015). mean using “dplyr v. 1.0.7” (Wickham et al., 2021).
Phenotypic plasticity variables were determined by
To partition the relative effects of climate and the trait parent mean (F1) under control
habitat fragmentation on trait variation in response environmental conditions minus the trait parent
to drought, we separately modelled each mean under drought stress conditions. The
explanatory variable and their two-way interaction redundancy analysis was modelled with the traits

8
scaled to unit variance as response and the climate southern populations (Fig. 2; Appendix S4). When
PC1 and habitat fragmentation PC1 as explanatory seeds were not vernalised, northern populations
variables. Furthermore, phenotypic trait variation germinated significantly slower. In general, the
and phenotypic plasticity under drought stress germination rate was significantly higher in
variables were evaluated with cluster analysis. A southern populations compared to northern
dissimilarity matrix was calculated using “cluster populations. As expected, juvenile growth (PC1)
v.1.2.1” (Maechler et al., 2017) with gower was significantly faster in southern populations
distances (Gower, 1971) and distinct clusters were compared to northern populations. Even though
delineated with hierarchical cluster analysis using drought stress only explained a minor fraction of
the ward.D criterion (Murtagh & Legendre, 2014) the explained variation, it significantly reduced
and a three group threshold (height = 1). Increasing juvenile growth rate in northern populations while it
the groups (or reducing the height) further led to decreased in southern connected populations
nested groups and reduced the overall ecological (large southern populations in contiguous habitat).
interpretability. Finally, the independent Flower phenology was significantly more sensitive
contribution of climate and habitat fragmentation to temperature cues (CCD4) in southern
was evaluated with variance partitioning using two populations compared to northern ones, which
partial redundancy analyses with respectively resulted in faster flower development (Fig. 2).
climate PC1 and habitat fragmentation PC1 as Under drought stress, northern populations
conditional variables. flowered significantly faster while for southern
populations flower development slowed down.
3. Results During the adult life phase, most of the explained
There was considerable climate driven trait variation in growth rate was driven by suppressed
variation along the range (O1). Vernalised northern growth under drought stress and this was most
seeds responded faster to temperature (cumulative pronounced in southern populations. The root-
degree days warmer than 4°C) compared to shoot ratio steeply increased in southern

Figure 2: The three-way interaction effect of climate, habitat fragmentation and drought stress frequency (or vernalisation during
germination) on early life phases and phenological processes of Primula elatior. The germination lag time was expressed as cumulative
degree days warmer than 4°C. The germination rate, here expressed as the germination percentage, is based on the germinated count
relative to the total seeds (model weights) per petri dish as determined by the quasibinomial model. Juvenile growth rate displays the PC
eigenvalues of the dimensionally reduced and imputed seedling growth rate, juvenile growth rate, and overall juvenile growth (from
seedling to adult). The flowering lag time as an indicator of phenology displays the cumulative degree days (>4°C).

9
connected populations, indicating an increased However, the discrepancy between northern and
investment in below-ground biomass (Fig. 3). southern populations decreased under drought
However, under drought stress, we observed a stress.
decrease in the root-shoot ratio in individuals
originating from southern populations and an The abaxial stomatal density was not significantly
increase in those from northern populations, influenced by climate along the range (Fig. 4).
specifically in landscapes unaffected by Under droughts stress, however, southern
fragmentation (connected). The specific leaf area populations increased the abaxial stomatal density
was significantly lower in southern populations. while it decreased in northern populations. Adaxial
stomatal density, on the other hand, did not display
phenotypic plasticity and linearly decreased
towards the south. The abaxial trichome density
was slightly higher in northern populations and this
increased further under drought stress.
Habitat fragmentation interfered with climatic clines
(O2) for several traits (i.e. significant two-way
interactions between habitat fragmentation and
climate), and for most drought responses (i.e. two-
way interactions between fragmentation and
drought stress; and the three-way interactions
between climate, habitat fragmentation and
drought stress). Specifically, we found that the
juvenile plant growth rate was relatively stable
across the range for connected populations but
was lower for fragmented populations in the north
and higher in the south of Primula elatior’s
distribution range. However, it has to be noted that
even though the interaction between climate and
habitat fragmentation was highly significant, the
partial contribution of the interaction effect to the
explained variance of the model was lower than the
contribution of climate and habitat fragmentation
separately (hence the negative contribution).
The flowering lag time was higher in southern
fragmented populations, while northern
populations flowered significantly slower under
drought stress. Habitat fragmentation significantly
limited adult growth in southern populations and
Figure 3: The three-way interaction effect of climate, habitat increased adult growth in northern populations,
fragmentation and drought stress frequency on adult growth and
thus nullifying the climate cline. However, the
biomass allocation strategies of Primula elatior. Three-way
interactions were predicted with climate PC1 as continuous contribution to the explained variance was limited.
variable, habitat fragmentation as moderator variable using all The root-shoot ratio increased from north to south
observed values (ranging from red to green), and drought as an expected evolutionary response to
treatment as binary categorical variable. The venn diagrams warmer/drier conditions, but this adaptive response
display the standardised contribution of fixed explained variance, was absent across fragmented populations.
as determined in the variance partitioning. The adult growth rate Moreover, the plasticity response of the root:shoot
displays the growth in mm per day from 83 days up to 337 days of
ratio were inverted in plants from fragmented
plant development. The biomass allocation displays the ratio
between the dried root biomass divided by the dried shoot biomass populations compared to plants from connected
in grams. The specific leaf area displays the leaf area divided by the populations. Phenotypes from connected
leaf mass in mm²/mg. populations evolved higher plasticity towards

10
drought stress with respect to abaxial stomatal which 16.27% was explained by climate (climate
densities, but this adaptive drought response was PC1 | habitat fragmentation PC1), 1.13% by habitat
absent across fragmented populations. fragmentation (habitat fragmentation PC1 | climate
Additionally, fragmented populations displayed a PC1), and 0.33% of the variation could not be
reduced abaxial glandular trichome density in separated between both drivers. Climate was
southern fragmented populations. strongly aligned with the dissimilarity-based trait
distances with plasticity of the flowering lag time
Flower investment was strongly reduced in and juvenile growth rate as strongest outliers (Fig.
fragmented populations along the range, being 6). Even though there was considerable
most pronounced in southern populations (Fig. 5). dissimilarity between the individuals with the same
Pollinator signalling (bee view), herkogamy, and parental origin (F1) along fragmentation levels, trait
selfing success was not primarily driven by habitat dissimilarity was considerably less with lowering
fragmentation (O3), but rather showed substantial investment (V) as strongest outlier.
north to south variation (Fig. 5). The first PC values
of the bee view increased in the south of the
distribution and this adaptation was significantly
more pronounced in fragmented habitats, which
indicates an overall higher median flower
reflectance of colour bands most visible to the
clade Anthophila (blue, green, UV red). However,
the large differentiation of flower reflectance
between fragmentation levels was not apparent in
northern populations. In addition, S morphs had a
significantly higher overall reflectance than L
morphs. Herkogamy, as an indicator of
heteromorphic self-incompatibility, significantly
increased in the south and this effect was strongest
in S morphs from connected populations. However,
S morphs originating from fragmented populations
had a significantly lower herkogamy over the whole
range and the divergence was especially
pronounced among southern populations.
Biochemical self-compatibility was significantly
higher in northern S-morphs close to the
distribution edge, and this effect was most
pronounced in fragmented populations. Succesful
self-pollination and increased inbreeding
coefficients in resulting offspring as compared to
outcrossed offspring pointed to a successful
breakdown of biochemical self-compatibility
(Appendix S5). However, even though habitat
fragmentation explained a high amount of variation
(34.21%), it was not possible to discern a statistical
Figure 4: Three-way interactions of climate, habitat fragmentation
trend due to a high variability between populations and drought stress on the abaxial and adaxial stomatal density,
and also because connected populations at the and the abaxial trichome densities of Primula elatior. Three-way
northern edge of the range displayed an overall interactions were predicted with climate PC1 as continuous
higher selfing rate. variable, habitat fragmentation as moderator variable using all
observed values (ranging from red to green), drought treatment
Climate (F = 86.77, p = 0.001) and habitat as binary categorical variable. The venn diagrams display the
fragmentation (F = 6.83, p = 0.001) explained standardised contribution of fixed explained variance, as
17.73% of the total phenotypic trait variation of determined in the variance partitioning.

11
 Figure 5: Three-way interactions of climate, habitat fragmentation and drought stress on flower investment, bee view, herkogamy, and
bio-chemical self-incompatibility of Primula elatior. The three-way interaction was predicted with climate PC1 as continuous variable,
habitat fragmentation as moderator variable using all observed values (ranging from red to green), and drought treatment as binary
categorical variable. Flower investment displays the first PC axis of the imputed total count of produced flowers, the mean flower
duration (of three stalks), total count of produced stalks, and the average stalk height. The bee view displays the first PC axis of the
median flower colour value (0-255) of UV red, green and blue bands, with higher values representing increased reflectance. Herkogamy
was based on a separate large field measurement dataset (60 populations) and displays the absolute distance between anther and
stigma. Self-incompatibility displays the relative percentage of selfed flowers that produced seeds. The venn diagrams display the
standardised contribution of fixed explained variance, as determined in the variance partitioning.

hormone signalling pathways responsible for

4. Discussion environment-dependent germination (Finch-
4.1. Climate adaptation Savage & Leubner-Metzger, 2006). A decrease in
Our results demonstrate considerable climate seed dormancy in southern populations can help
adaptation across the range of P. elatior in distinct minimize trade-offs associated with dormancy
life phases and traits (Fig. 6), which in turn while enhancing fitness across various life stages
determines plant performance and population (Postma & Ågren, 2022). The strong climate cline
viability (Lortie & Hierro, 2022). The potential for in the germination rate is consistent with previous
early plant performance is primarily driven by seed analyses of slow-colonizing forest herb species
dormancy, lag time (initiation), germination where an increased germination success in
success, and early growth strategies. Here, we warmer climates was observed (De Frenne et al.,
found that seeds from northern populations were 2011; Frenne et al., 2012; Graae et al., 2009).
more sensitive to vernalisation (less cumulative Furthermore, warm-adapted progeny enhanced
degree days, warmer than 4°C, before their juvenile growth rates under drought stress,
germination) compared to southern populations whereas cold-adapted progeny exhibited reduced
(Fig. 2), indicating evolution towards increased juvenile growth rates under drought. This suggests
seed dormancy in colder climates in order to avoid that warm-adapted progeny, which evolved to
frost damage while maximizing the growing season optimize growth rates in warmer climates (Chaves
(Bewley et al., 2012). The observed differentiation et al., 2003; Morison & Morecroft, 2008), display
in seed dormancy across the range can therefore increased drought tolerance (Kooyers, 2015).
be considered as a consequence of natural Flowering was significantly delayed in northern
selection in key regulatory genes that manage the populations, which is consistent with the general
trend that temperate forest communities tend to

12
 Figure 6: Redundancy analysis of phenotypic trait variation (Adaptation) and phenotypic plasticity (Plasticity) in Primula elatior. Traits
were scaled to unit variance as response, and the first PC axis of respectively climate and habitat fragmentation were used as
explanatory variables. Each dot indicates a parental origin ID (individuals grouped by F1), colours are defined by the first climate PC
axis eigenvalues, the dot size is determined by the eigenvalues of the first habitat fragmentation PC axis. The x axis displays the
eigenvalues of the first RDA axis and the y axis displays the eigenvalues of the second RDA axis. The legend for the trait IDs is displayed
on the right. The differentiation between phenotypic trait variation (Adaptation) and phenotypic plasticity (Plasticity) is indicated in the
response column.
flower later in colder climates to prevent frost generally display distinct plastic responses to cope
damage (Flynn & Wolkovich, 2018) and to with drought stress, aiming to either avoid, escape,
synchronize their blooming with the emergence of or tolerate drought (Kooyers, 2015). In this study,
pollinator communities (Memmott et al., 2007). adult growth slowed down, the root:shoot ratio
Under drought stress, southern populations further decreased, SLA increased, and abaxial stomata
slowed flowering, which is in line with the increased under drought stress. Slowing down
observation that the flowering season of vernal adult growth indicates a drought avoidance
herbs lag behind under reduced precipitation in strategy aiming to reduce water loss (Chaves et al.,
order to avoid unsuccessful fruit development 2003; Whale, 1983). A reduced root:shoot ratio and
under dry conditions (Nam & Kim, 2020). increased specific leaf area under drought may
initially appear counterintuitive, however, by
The overall higher adult growth rate, root:shoot allocating more resources to shoots and expanding
ratio and lower specific leaf area of southern leaf area, plants can efficiently maintain
populations is in line with the general observation photosynthesis in water-limited conditions, a
that herbaceous plants increase relative characteristic often observed in shade-tolerant
investment in root biomass and increase leaf species (McCarthy & Enquist, 2007; Poorter et al.,
density to reduce evaporation and enhance carbon 2012). Increased abaxial stomatal density enables
assimilation in warmer climates (Frenne et al., plants to enhance their capacity for gas exchange
2012; Lemke et al., 2015; Poorter et al., 2009). regulation, which in turn increases their carbon
Plants, depending on their evolutionary history, will assimilation rate and water use efficiency while

13
maintaining photosynthesis (Xu & Zhou, 2008). 2017; Hadley & Betts, 2012; Naaf et al., 2021). The
The adaptive drought avoidance strategy of resulting genetic erosion likely led to the observed
southern populations enabled plants to maintain reduced adaptive potential in several key traits
their flowering investment under drought (Karlsson related to development and water regulation and
& Méndez, 2005). Conversely, northern could hinder their ability to cope with climate
populations exhibited an increased root:shoot change. The observed reduction of the glandular
ratio, along with a reduction in specific leaf area trichome density in southern fragmented
and abaxial stomatal density under drought stress, populations could be an additional consequence of
suggesting an adaptive drought tolerance strategy disrupted plant-pollinator networks, potentially
(Kooyers, 2015; Poorter et al., 2012). leading to a diminished investment in secondary
metabolites which attract pollinators and,
4.2. The impact of habitat fragmentation consequently, decreasing the adaptive advantage
on climate adaptation of glandular trichomes (Uzelac et al., 2012). As
In line with our second objective, we found that glandular trichomes and the produced secondary
habitat fragmentation disrupted climate clines and metabolites also play a role in protection against
adaptive responses to drought stress in several UV, drought, and predation this could thus further
traits and across life phases. Specifically, southern lead to a reduced stress tolerance (Colombo et al.,
fragmented populations of P. elatior exhibited 2017; Uzelac et al., 2012). Finally, fragmented
longer flowering lag times, the adult growth cline habitats often experience increased edge effects,
was absent in fragmented populations, the relation leading to altered microclimates (Harper et al.,
between climate and the root:shoot ratio was 2005; Hofmeister et al., 2019). Heterogeneous
inverted in fragmented populations, and the adaxial landscapes with a high fragmentation rate also
trichome density was strongly reduced in southern tend to contain more early-successional secondary
fragmented populations, leading to strongly forests, characterized by distinct tree communities
reduced flower investment in fragmented with unique phenology (Dubois & Cheptou, 2017;
populations. Additionally, in fragmented Vandepitte et al., 2007). Furthermore, small forest
populations, the adaptive capacity to drought fragments often exhibit elevated soil nutrient
stress was impacted across various traits and life concentrations due to leaching from surrounding
stages. Phenotypic plasticity of the growth rate fields, leading to increased aboveground
during both juvenile and adult phases, as well as competition (Weathers et al., 2001). These
the root:shoot ratio, were reduced in these differentiated selection pressures could have led to
populations, and plasticity in abaxial stomatal the slightly faster germination(lag time) and the
density was completely lacking. increased juvenile growth rate enabling
populations to occasionally establish in more
The effective population size plays a crucial role in recent forest stands where interspecific
shaping adaptive differentiation by influencing the competition is more prevalent (Dubois & Cheptou,
interplay between selection and genetic drift. 2017; Hermy et al., 1999). Furthermore, phenology
Larger populations are more likely to maintain a of P. elatior is highly sensitive to environmental
higher genetic diversity and experience stronger cues and the observed delayed flowering could
selection pressures, which in turn fosters adaptive disrupt plant-pollinator networks in fragmented
differentiation in response to varying climatic forests (Baeten et al., 2015). In conclusion, habitat
conditions (Leimu & Fischer, 2008). On the other fragmentation disrupts climate clines and adaptive
hand, smaller populations are more susceptible to responses to drought stress in P. elatior, potentially
genetic drift, resulting in the random loss of reducing the long-term resilience and viability of
advantageous alleles (Aguilar et al., 2008; Naaf et populations. Our study highlights the importance of
al., 2021; Pickup et al., 2012). Additionally, habitat understanding the complex interplay between
fragmentation can lead to reduced gene flow habitat fragmentation and species' adaptive
between populations due to an unsuitable potential in the face of ongoing environmental
landscape matrix and an altered or degraded changes. Further untangling the evolutionary
pollinator community, further reducing genetic factors contributing to diminished climate clines in
diversity and evolutionary potential (Cheptou et al., fragmented habitats will be vital for future research

14
endeavours. Considering habitat fragmentation and underwent strong selection pressures in often
when evaluating local adaptation not only isolated founding populations (Koski et al., 2019),
enhances our understanding of how complex leading to a decreased herkogamy and a higher
environmental pressures shape adaptive self-compatibility in northern populations (Barrett,
processes but also ensures accurate guidance for 2019; Encinas-Viso et al., 2020). Long-term
conservation efforts aimed at effectively preserving isolation in glacial refugia and the northward post-
biodiversity. glacial recolonization of Europe thus likely exerted
strong evolutionary pressures on herkogamy
4.3. The role of habitat fragmentation in differentiation over extended periods of time (Cain
shaping mating systems et al., 1998; Cheddadi et al., 2006), but
Habitat fragmentation has a notable impact on the contemporary habitat fragmentation is already
mating system of Primula elatior, affecting leaving its mark on the breakdown of herkogamy in
evolutionary selection in multiple flower traits (O3). Primula elatior. S-morphs, being more susceptible
Flower investment was strongly reduced in to gravity or wind-induced self-pollination,
fragmented populations along the range, experience increased selection pressures on
especially in southern populations (Fig. 5). herkogamy breakdown, making these effects
Additionally, southern populations exhibited particularly evident in these morphs. Attraction
divergent pollinator signalling (bee view) between mechanisms, such as bullseye patterns and UV
fragmented and connected habitats. In these signalling, can enhance pollinator orientation, and
southern fragmented populations, S morphs also the reduced pollinator availability in fragmented
demonstrated a marked reduction in herkogamy. habitats could therefore exert selective pressure on
Furthermore, the presence of biochemical self- flowers with increased pollinator signalling (Chittka
compatibility in fragmented habitats led to a & Wells, 2004; Koski & Ashman, 2014). Pollinator
substantial increase in inbreeding coefficients signalling was mainly observed in southern
among selfed offspring (Appendix S5). As flowers fragmented populations, which may be linked to
in fragmented populations likely receive less flower the biogeographical history along the range or to
visits from pollinators, investment in flowers climate-driven gradients in the pollinator
becomes less beneficial from an evolutionary community (Ganuza et al., 2022; Van Daele et al.,
viewpoint (Aguilar et al., 2006). Nevertheless, it 2022). Future phylogenetically-controlled analyses
remains uncertain whether the diminished flower could help discern how the increased reflection
investment stems from disrupted plant-pollinator (higher PC1 values of median UV red, green and
networks (Jacquemyn et al., 2012), reduced blue bands) in fragmented southern populations
climate clines in fragmented populations (section affects pollinators and how this related to the
4.2), from inbreeding effects (Andersson, 2012), limited or altered pollinator community in these
reduced fitness (Lienert, 2004) or other habitats. In summary, habitat fragmentation
environmental drivers (Weber & Kolb, 2011). considerably influences the evolutionary trajectory
of P. elatior through its effects on mating system
While flower investment was mainly associated attributes such as reduced flower investment,
with fragmentation, herkogamy, pollinator altered herkogamy, increased inbreeding
signalling (bee view), and self-compatibility coefficients among selfed offspring, and evolving
displayed substantial north to south variation. flower attraction patterns. Further exploration of
During the last ice age, the ranges of many forest these intricate relationships will be vital for
herbs retracted, with remaining populations informing effective conservation practices.
residing in temperate regions throughout the Last
Glacial Maximum. Postglacial recolonization by
dispersal limited forest herbs was particularly slow, 5. Conclusion and implications
still impacting species distributions and genetic Primula elatior demonstrated significant climate
imprints to date (Hewitt, 1999; Svenning et al., adaptation across life phases, with southern
2008; Willner et al., 2009). During the northward populations displaying a drought avoidance
post-glacial recolonization of Europe many forest strategy and northern populations displaying a
herb populations accumulated genetic bottlenecks drought tolerance strategy. These adaptations

15
enable populations to cope with their specific practical support. Furthermore, we would like to
climatic conditions, ensuring their long-term thank TRANSfarm and Meise Botanic Garden for
survival and viability. However, habitat the use of facilities. Finally, we would like to thank
fragmentation poses a significant threat to species' Prof. Francis Wyffels for useful comments in
climate adaptation and resilience by disrupting regards to the training of the neural networks for
adaptive responses to drought stress and altering the detection of stomata and glandular trichomes.
plant strategies. Fragmented populations showed
reduced phenotypic plasticity in key traits related to 7. Author Contributions
development and water regulation, potentially Frederik Van Daele conducted the fieldwork (seed
hindering their ability to cope with climate change.
collection and herkogamy data), set up the common
Furthermore, habitat fragmentation significantly
garden, measured trait data, developed the image
impacted the mating system of P. elatior, severely
affecting flower investment. The long-term recognition tools, performed the statistical analysis,
resilience and viability of P. elatior populations made the figures, and wrote the initial draft of the
could be reduced due to eco-evolutionary effects of manuscript. Olivier Honnay provided feedback on the
habitat fragmentation on climate adaptation. Our experimental set-up and on the initial draft. Steven
findings have important implications for Janssens supervised microscopy of stomata and
conservation management and planning. trichomes and provided feedback on the initial draft.
Understanding the complex interplay between Hanne De Kort conceived the experiment, advised the
habitat fragmentation and species' adaptive experimental set-up, and provided feedback on the
potential is crucial for developing effective initial draft.
strategies to protect and maintain biodiversity in
the face of ongoing environmental changes.
Conservation efforts should focus on restoring and
8. Data Availability Statement
Trait data will be deposited in the TRY Plant Trait
preserving large, connected habitats, which can
enhance gene flow and maintain the genetic Database upon publication of the manuscript. A direct
diversity necessary for species to adapt to link to these datasets will be provided in the final
changing environmental conditions. Furthermore, article.
considering habitat fragmentation when evaluating
local adaptation in progeny is essential for 9. Conflict of interest
accurately guiding conservation efforts and The authors declare no conflict of interest.
ensuring long-term species survival. Future
research could further explore the evolutionary
drivers of reduced climate clines in fragmented
10. References
habitats, as well as investigating the effects of Abadi, M., Agarwal, A., Barham, P., Brevdo, E.,
habitat fragmentation on other plant species and Chen, Z., Citro, C., Corrado, G. S., Davis, A.,
their capacity to adapt to environmental change. Dean, J., Devin, M., Ghemawat, S.,
Addressing these knowledge gaps will enable Goodfellow, I., Harp, A., Irving, G., Isard, M.,
more targeted and effective conservation Jia, Y., Jozefowicz, R., Kaiser, L., Kudlur, M.,
strategies to be developed, ultimately contributing … Zheng, X. (2015). TensorFlow: Large-
to the preservation of global biodiversity in a rapidly Scale Machine Learning on Heterogeneous
changing world. Systems. https://www.tensorflow.org/
Aguilar, R., Ashworth, L., Galetto, L., & Aizen, M.
6. Acknowledgments A. (2006). Plant reproductive susceptibility to
This research was funded by the Flemish habitat fragmentation: Review and synthesis
Research Foundation (FWO project G091419N). through a meta-analysis. Ecology Letters,
We would like to thank Kasper Van Acker, Jonas 9(8), 968–980. https://doi.org/10.1111/j.1461-
Lequeu, Lander Storms, Olivia De Paepe, Niels 0248.2006.00927.x
Duym, Emily Kirsch, Gilles Breugelmans, Jarne
Aguilar, R., Quesada, M., Ashworth, L., Herrerias-
Melis, Katlijne Nachtergaele, and Lotte Simons for
Diego, Y., & Lobo, J. (2008). Genetic

16
 consequences of habitat fragmentation in (2015). Fitting Linear Mixed-Effects Models
plant populations: Susceptible signals in Using {lme4}. Journal of Statistical Software,
plant traits and methodological approaches. 67(1), 1–48.
Molecular Ecology, 17(24), 5177–5188. https://doi.org/10.18637/jss.v067.i01
https://doi.org/10.1111/j.1365-
294X.2008.03971.x Bawa, K. . S. ., Brys, R., de Crop, E., Hoffmann,
M., Jacquemyn, H., Busch, J. W., Schoen, D.
Aiken, L. S., & West, S. G. (1991). Multiple J., Hiscock, S. . J., Lloyd, D. G. ., Mansur, L.,
regression: testing and interpreting Gonzalez, M., Rojas, I., Salas, P., Ortiz, M.
interactions. Sage. A., Talavera, S., García-Castaño, J. L.,
Tremetsberger, K., Stuessy, T., Balao, F., …
Andersson, S. (2012). Does inbreeding promote Tabah, D. a. (2011). Mechanisms of self-
evolutionary reduction of flower size? incompatibility in flowering plants. American
experimental evidence from Crepis tectorum Journal of Botany, 66(1), 1988–2007.
(Asteraceae). American Journal of Botany, https://doi.org/10.1111/j.1558-
99(8), 1388–1398. 5646.2011.01505.x
https://doi.org/10.3732/ajb.1200116
Becklin, K. M., Anderson, J. T., Gerhart, L. M.,
Armbruster, W. S. (2002). Can indirect selection Wadgymar, S. M., Wessinger, C. A., & Ward,
and genetic context contribute to trait J. K. (2016). Examining plant physiological
diversification? A transition-probability study responses to climate change through an
of blossom-colour evolution in two genera. evolutionary lens. Plant Physiology, 172(2),
Journal of Evolutionary Biology, 15(3), 468– 635–649.
486. https://doi.org/10.1046/j.1420- https://doi.org/10.1104/pp.16.00793
9101.2002.00399.x
Bewley, J. D., Bradford, K., & Hilhorst, H. (2012).
Baeten, L., Sercu, B., Bonte, D., Vanhellemont, Dormancy and the Control of Germination. In
M., & Verheyen, K. (2015). Intraspecific Seeds: Physiology of Development,
variation in flowering phenology affects seed Germination and Dormancy (pp. 247–279).
germinability in the forest herb Primula Springer Science \& Business Media.
elatior. Plant Ecology and Evolution, 148(2),
283–288. Blanquart, F., Kaltz, O., Nuismer, S. L., &
https://doi.org/10.5091/plecevo.2015.1089 Gandon, S. (2013). A practical guide to
measuring local adaptation. Ecology Letters,
Barrett, S. C. H. (2019). ‘A most complex 16(9), 1195–1205.
marriage arrangement’: recent advances on https://doi.org/10.1111/ele.12150
heterostyly and unresolved questions. New
Phytologist, 224(3), 1051–1067. Bonte, D., Van Dyck, H., Bullock, J. M., Coulon,
https://doi.org/10.1111/nph.16026 A., Delgado, M., Gibbs, M., Lehouck, V.,
Matthysen, E., Mustin, K., Saastamoinen, M.,
Barthelme, S. (2022). imager: Image Processing Schtickzelle, N., Stevens, V. M.,
Library Based on “CImg.” https://cran.r- Vandewoestijne, S., Baguette, M., Barton, K.,
project.org/package=imager Benton, T. G., Chaput-Bardy, A., Clobert, J.,
Baskin, C. C. (2014). Seeds: Ecology, Dytham, C., … Travis, J. M. J. (2012). Costs
Biogeography, And, Evolution of Dormancy of dispersal. Biological Reviews, 87(2), 290–
and Germination (2nd ed.). Academic Press. 312. https://doi.org/10.1111/j.1469-
185X.2011.00201.x
Baskin, J. M., & Baskin, C. C. (2004). A
classification system for seed dormancy. Brys, R., & Jacquemyn, H. (2015). Disruption of
Seed Science Research, 14(July), 1–16. the distylous syndrome in Primula veris.
https://doi.org/10.1079/SSR2003150 Annals of Botany, 115(1), 27–39.
https://doi.org/10.1093/aob/mcu211
Bates, D., Mächler, M., Bolker, B., & Walker, S.

17
Bucher, S. F., König, P., Menzel, A., Migliavacca, Discrete Event Simulation Models.
M., Ewald, J., & Römermann, C. (2018). https://cran.r-project.org/package=SpaDES
Traits and climate are associated with first
flowering day in herbaceous species along Cleland, E. E., Isabelle, C., & Annette, M. (2007).
elevational gradients. Ecology and Evolution, Shifing Plant Phenology In Response to
8(2), 1147–1158. Global Change. Trends in Ecology &
https://doi.org/10.1002/ece3.3720 Evolution, 22(7), 357–365.

Cain, M. L., Damman, H., & Muir, A. (1998). Seed Colombo, P. S., Flamini, G., Rodondi, G., Giuliani,
dispersal and the Holocene migration of C., Santagostini, L., & Fico, G. (2017).
woodland herbs. Ecological Monographs, Phytochemistry of European Primula
68(3), 325–347. species. Phytochemistry, 143, 132–144.
https://doi.org/10.1890/0012- https://doi.org/10.1016/j.phytochem.2017.07.
9615(1998)068[0325:SDATHM]2.0.CO;2 005

Chaves, M. M., Maroco, J. P., & Pereira, J. S. Coors, A., & DeMeester, L. (2009). Synergistic,
(2003). Understanding plant responses to antagonistic and additive effects of multiple
drought - From genes to the whole plant. stressors: Predation threat, parasitism and
Functional Plant Biology, 30(3), 239–264. pesticide exposure in Daphnia magna.
https://doi.org/10.1071/FP02076 Journal of Applied Ecology, 46(5), 1138.
https://doi.org/10.1111/j.1365-
Cheddadi, R., Vendramin, G. G., Litt, T., François, 2664.2009.01727-1.x
L., Kageyama, M., Lorentz, S., Laurent, J.
M., de Beaulieu, J. L., Sadori, L., Jost, A., & Copernicus Land Monitoring Service. (2018).
Lunt, D. (2006). Imprints of glacial refugia in Forest Type (HRL_ForestType_2018).
the modern genetic diversity of Pinus European Environment Agency.
sylvestris. Global Ecology and Biogeography, https://land.copernicus.eu/pan-
15(3), 271–282. european/high-resolution-
https://doi.org/10.1111/j.1466- layers/forests/forest-type-1/status-
822X.2006.00226.x maps/forest-type-2018?tab=mapview

Chen, H. (2018). VennDiagram: Generate High- Cote, J., Bestion, E., Jacob, S., Travis, J.,
Resolution Venn and Euler Plots. Legrand, D., & Baguette, M. (2017).
https://cran.r- Evolution of dispersal strategies and
project.org/package=VennDiagram dispersal syndromes in fragmented
landscapes. Ecography, 40(1), 56–73.
Cheptou, P. O., Hargreaves, A. L., Bonte, D., & https://doi.org/10.1111/ecog.02538
Jacquemyn, H. (2017). Adaptation to
fragmentation: Evolutionary dynamics driven Dawson, J. F., & Richter, A. W. (2006). Probing
by human influences. Philosophical three-way interactions in moderated multiple
Transactions of the Royal Society B: regression: Development and application of a
Biological Sciences, 372(1712). slope difference test. Journal of Applied
https://doi.org/10.1098/rstb.2016.0037 Psychology, 91(4), 917–926.
https://doi.org/10.1037/0021-9010.91.4.917
Chittka, L., & Wells, H. (2004). Color vision in
bees: mechanisms, ecology, and evolution. De Frenne, P., Brunet, J., Shevtsova, A., Kolb, A.,
How Simple Nervous Systems Create Graae, B. J., Chabrerie, O., Cousins, S. A.,
Complex Perceptual Worlds, 165–191. Decocq, G., De Schrijver, A., Diekmann, M.,
Gruwez, R., Heinken, T., Hermy, M., Nilsson,
Chollet, F., & others. (2015). Keras. GitHub. C., Stanton, S., Tack, W., Willaert, J., &
https://github.com/fchollet/keras Verheyen, K. (2011). Temperature effects on
forest herbs assessed by warming and
Chubaty, A. M., & McIntire, E. J. B. (2022). transplant experiments along a latitudinal
SpaDES: Develop and Run Spatially Explicit

18
 gradient. Global Change Biology, 17(10), 241. https://doi.org/10.1111/j.0906-
3240–3253. https://doi.org/10.1111/j.1365- 7590.2004.03731.x
2486.2011.02449.x
Eriksson, O., & Ehrlen, J. (2001). Landscape
De Frenne, P., Graae, B. J., Rodríguez-Sánchez, fragmentation and the viability of plant
F., Kolb, A., Chabrerie, O., Decocq, G., De populations (J. Silvertown (ed.)). Blackwell.
Kort, H., De Schrijver, A., Diekmann, M.,
Eriksson, O., Gruwez, R., Hermy, M., Lenoir, Etterson, J. R., & Shaw, R. G. (2001). Constraint
J., Plue, J., Coomes, D. A., & Verheyen, K. to adaptive evolution in response to global
(2013). Latitudinal gradients as natural warming. Science, 294(5540), 151–154.
laboratories to infer species’ responses to https://doi.org/10.1126/science.1063656
temperature. Journal of Ecology, 101(3), Fick, S. E., & Hijmans, R. J. (2017). WorldClim 2:
784–795. https://doi.org/10.1111/1365- new 1-km spatial resolution climate surfaces
2745.12074 for global land areas. International Journal of
De Kort, H., Panis, B., Helsen, K., Douzet, R., Climatology, 37(12), 4302–4315.
Janssens, S. B., & Honnay, O. (2020). Pre- https://doi.org/10.1002/joc.5086
adaptation to climate change through Finch-Savage, W. E., & Leubner-Metzger, G.
topography-driven phenotypic plasticity. (2006). Seed dormancy and the control of
Journal of Ecology, 108(4), 1465–1474. germination. In New Phytologist (Vol. 171,
https://doi.org/10.1111/1365-2745.13365 Issue 3, pp. 501–523).
De Witte, L. C., & Stöcklin, J. (2010). Longevity of https://doi.org/10.1111/j.1469-
clonal plants: Why it matters and how to 8137.2006.01787.x
measure it. Annals of Botany, 106(6), 859– Flanders Research Foundation. (2022). Flemish
870. https://doi.org/10.1093/aob/mcq191 Supercomputer Center.
Dickinson, H. G. (1990). Self‐incompatibility in https://www.vscentrum.be
flowering plants: Evolution, Diversity, and Flynn, D. F. B., & Wolkovich, E. M. (2018).
Mechanisms (V. E. Franklin-Tong (ed.); Vol. Temperature and photoperiod drive spring
12, Issue 4). Springer. phenology across all species in a temperate
https://doi.org/10.1002/bies.950120403 forest community. New Phytologist, 219(4),
Dubois, J., & Cheptou, P. O. (2017). Effects of 1353–1362.
fragmentation on plant adaptation to urban https://doi.org/10.1111/nph.15232
environments. Philosophical Transactions of Franks, S. J., Weber, J. J., & Aitken, S. N. (2014).
the Royal Society B: Biological Sciences, Evolutionary and plastic responses to climate
372(1712). change in terrestrial plant populations.
https://doi.org/10.1098/rstb.2016.0038 Evolutionary Applications, 7(1), 123–139.
Dunnington, D., & Harvey, P. (2021). exifr: EXIF https://doi.org/10.1111/eva.12112
Image Data in R. https://cran.r- Frenne, P. De, Graae, B. J., Brunet, J.,
project.org/package=exifr Shevtsova, A., Schrijver, A. De, Chabrerie,
Encinas-Viso, F., Young, A. G., & Pannell, J. R. O., Cousins, S. A. O., Decocq, G.,
(2020). The loss of self-incompatibility in a Diekmann, M., Hermy, M., Heinken, T., Kolb,
range expansion. Journal of Evolutionary A., Nilsson, C., Stanton, S., & Verheyen, K.
Biology, 33(9), 1235–1244. (2012). The response of forest plant
https://doi.org/10.1111/jeb.13665 regeneration to temperature variation along a
latitudinal gradient. Annals of Botany, 109(5),
Endels, P., Adriaens, D., Verheyen, K., & Hermy, 1037–1046.
M. (2004). Population structure and adult https://doi.org/10.1093/aob/mcs015
plant performance of forest herbs in three
contrasting habitats. Ecography, 27(2), 225– Ganuza, C., Redlich, S., Uhler, J., Tobisch, C.,

19
 Rojas-Botero, S., Peters, M. K., Zhang, J., Harper, K. A., Macdonald, S. E., Burton, P. J.,
Benjamin, C. S., Englmeier, J., Ewald, J., Chen, J., Brosofske, K. D., Saunders, S. C.,
Fricke, U., Haensel, M., Kollmann, J., Riebl, Euskirchen, E. S., Roberts, D., Jaiteh, M. S.,
R., Uphus, L., Müller, J., & Steffan-Dewenter, & Esseen, P. A. (2005). Edge influence on
I. (2022). Interactive effects of climate and forest structure and composition in
land use on pollinator diversity differ among fragmented landscapes. In Conservation
taxa and scales. Science Advances, 8(18). Biology (Vol. 19, Issue 3, pp. 768–782).
https://doi.org/10.1126/sciadv.abm9359 https://doi.org/10.1111/j.1523-
1739.2005.00045.x
Gómez-Martínez, C., Aase, A. L. T. O., Totland,
Ø., Rodríguez-Pérez, J., Birkemoe, T., Hartig, F. (2022). DHARMa: Residual Diagnostics
Sverdrup-Thygeson, A., & Lázaro, A. (2020). for Hierarchical (Multi-Level / Mixed)
Forest fragmentation modifies the Regression Models. https://cran.r-
composition of bumblebee communities and project.org/package=DHARMa
modulates their trophic and competitive
interactions for pollination. Scientific Reports, He, H. S., DeZonia, B. E., & Mladenoff, D. J.
10(1), 1–15. https://doi.org/10.1038/s41598- (2001). An aggregation index (AI) to quantify
020-67447-y spatial patterns of landscapes. Landscape
Ecology, 16(1), 87.
Gower, J. C. (1971). A General Coefficient of https://doi.org/10.1023/A:1017308405507
Similarity and Some of Its Properties.
Biometrics, 27(4), 857. Herlihy, C. R., & Eckert, C. G. (2007).
https://doi.org/10.2307/2528823 Evolutionary analysis of a key floral trait in
Aquilegia canadensis (Ranunculaceae):
Graae, B. J., Verheyen, K., Kolb, A., Van Der Genetic variation in herkogamy and its effect
Veken, S., Heinken, T., Chabrerie, O., on the mating system. Evolution, 61(7),
Diekmann, M., Valtinat, K., Zindel, R., 1661–1674. https://doi.org/10.1111/j.1558-
Karlsson, E., Ström, L., Decocq, G., Hermy, 5646.2007.00137.x
M., & Baskin, C. C. (2009). Germination
requirements and seed mass of slow- And Hermy, M., Honnay, O., Firbank, L., Grashof-
fast colonizing temperate forest herbs along Bokdam, C., & Lawesson, J. E. (1999). An
a latitudinal gradient. Ecoscience, 16(2), ecological comparison between ancient and
248–257. https://doi.org/10.2980/16-2-3234 other forest plant species of Europe, and the
implications for forest conservation.
Gumbert, A. (2000). Color choices by bumble Biological Conservation, 91(1), 9–22.
bees (Bombus terrestris): innate preferences https://doi.org/10.1016/S0006-
and generalization after learning. Behavioral 3207(99)00045-2
Ecology and Sociobiology, 48(1), 36–43.
http://link.springer.com/10.1007/s002650000 Hesselbarth, M. H. K., Sciaini, M., With, K. A.,
213 Wiegand, K., & Nowosad, J. (2019).
landscapemetrics : an open‐source R tool to
Hadley, A. S., & Betts, M. G. (2012). The effects calculate landscape metrics . Ecography, 1–
of landscape fragmentation on pollination 10. https://doi.org/10.1111/ecog.04617
dynamics: Absence of evidence not evidence
of absence. Biological Reviews, 87(3), 526– Hewitt, G. M. (1999). Post-glacial re-colonization
544. https://doi.org/10.1111/j.1469- of European biota. Biological Journal of the
185X.2011.00205.x Linnean Society, 68(1–2), 87–112.
https://doi.org/10.1006/bijl.1999.0332
Hanski, I., Mononen, T., & Ovaskainen, O. (2011).
Eco-evolutionary metapopulation dynamics Hijmans, R. J. (2019). raster: Geographic Data
and the spatial scale of adaptation. American Analysis and Modeling. R package version
Naturalist, 177(1), 29–43. 3.0-7. https://cran.r-
https://doi.org/10.1086/657625 project.org/package=raster

20
Hofmeister, J., Hošek, J., Brabec, M., Střalková, -allocation-in-plants/reekie/978-0-12-088386-
R., Mýlová, P., Bouda, M., Pettit, J. L., 8
Rydval, M., & Svoboda, M. (2019).
Microclimate edge effect in small fragments Katabuchi, & Masatoshi. (2015). LeafArea: An R
of temperate forests in the context of climate Package for Rapid Digital Image Analysis of
change. Forest Ecology and Management, Leaf Area. Ecological Research, 30(6),
448(May), 48–56. 1073–1077. https://doi.org/10.1007/s11284-
https://doi.org/10.1016/j.foreco.2019.05.069 015-1307-x

Honnay, O., Butaye, J., Jacquemyn, H., Kawecki, T. J., & Ebert, D. (2004). Conceptual
Verheyen, K., Hermy, M., Bossuyt, B., issues in local adaptation. Ecology Letters,
Butaye, J., Jacquemyn, H., Bossuyt, B., 7(12), 1225–1241.
Hermy, M., Verheyen, K., Hermy, M., & https://doi.org/10.1111/j.1461-
Bossuyt, B. (2002). Possible effects of 0248.2004.00684.x
habitat fragmentation and climate change on Keitt, T. H., Urban, D. L., & Milne, B. T. (1997).
the range of forest plant species. Ecology Detecting Critical Scales in Fragmented
Letters, 5(4), 525–530. Landscapes. Conservation Ecology, 1(1), 1–
https://doi.org/10.1046/j.1461- 13. https://eur-lex.europa.eu/legal-
0248.2002.00346.x content/PT/TXT/PDF/?uri=CELEX:32016R06
Jacquemyn, H., Brys, R., & Hermy, M. (2002). 79&from=PT%0Ahttp://eur-
Patch occupancy, population size and lex.europa.eu/LexUriServ/LexUriServ.do?uri
reproductive success of a forest herb =CELEX:52012PC0011:pt:NOT
(Primula elatior) in a fragmented landscape. Kooyers, N. J. (2015). The evolution of drought
Oecologia, 130(4), 617–625. escape and avoidance in natural herbaceous
https://doi.org/10.1007/s00442-001-0833-0 populations. Plant Science, 234, 155–162.
Jacquemyn, H., de Meester, L., Jongejans, E., & https://doi.org/10.1016/j.plantsci.2015.02.012
Honnay, O. (2012). Evolutionary changes in Koski, M. H., & Ashman, T. L. (2013). Quantitative
plant reproductive traits following habitat variation, heritability, and trait correlations for
fragmentation and their consequences for ultraviolet floral traits in Argentina anserina
population fitness. Journal of Ecology, (Rosaceae): Implications for floral evolution.
100(1), 76–87. https://doi.org/10.1111/j.1365- Evolution, 174(8), 1109–1120.
2745.2011.01919.x https://doi.org/10.1086/671803
Jaeger, J. A. G. (2000). Landscape division, Koski, M. H., & Ashman, T. L. (2014). Dissecting
splitting index, and effective mesh size: New pollinator responses to a ubiquitous
measures of landscape fragmentation. ultraviolet floral pattern in the wild. Functional
Landscape Ecology, 15(2), 115–130. Ecology, 28(4), 868–877.
https://doi.org/10.1023/A:1008129329289 https://doi.org/10.1111/1365-2435.12242
Jump, A. S., & Peñuelas, J. (2005). Running to Koski, M. H., Layman, N. C., Prior, C. J., Busch,
stand still: Adaptation and the response of J. W., & Galloway, L. F. (2019). Selfing ability
plants to rapid climate change. Ecology and drift load evolve with range expansion.
Letters, 8(9), 1010–1020. Evolution Letters, 3(5), 500–512.
https://doi.org/10.1111/j.1461- https://doi.org/10.1002/evl3.136
0248.2005.00796.x
Kwak, M. M., Velterop, O., & van Andel, J. (1998).
Karlsson, P. S., & Méndez, M. (2005). The Pollen and gene flow in fragmented habitats.
Resource Economy of Plant Reproduction. In Applied Vegetation Science, 1(1), 37–54.
E. Reekie & F. A. Bazzaz (Eds.), https://doi.org/10.2307/1479084
Reproductive Allocation in Plants. Elsevier.
https://www.elsevier.com/books/reproductive Lai, J., Zou, Y., Zhang, J., & Peres-Neto, P. R.

21
 (2022). Generalizing hierarchical and 110(5), 1015–1021.
variation partitioning in multiple regression https://doi.org/10.1111/1365-2745.13664
and canonical analyses using the rdacca.hp
R package. Methods in Ecology and Lüdecke, D. (2021). sjPlot: Data Visualization for
Evolution, 13(4), 782–788. Statistics in Social Science. https://cran.r-
https://doi.org/10.1111/2041-210X.13800 project.org/package=sjPlot

Leimu, R., & Fischer, M. (2008). A meta-analysis Lüdecke, D., Ben-Shachar, M. S., Patil, I.,
of local adaptation in plants. PLoS ONE, Waggoner, P., & Makowski, D. (2021).
3(12), 1–8. performance: An R Package for Assessment,
https://doi.org/10.1371/journal.pone.0004010 Comparison and Testing of Statistical
Models. Journal of Open Source Software,
Leimu, R., Vergeer, P., Angeloni, F., & Ouborg, N. 6(60), 3139.
J. (2010). Habitat fragmentation, climate https://doi.org/10.21105/joss.03139
change, and inbreeding in plants. Annals of
the New York Academy of Sciences, Lunau, K., Verhoeven, C., & Ren, Z.-X. (2018).
1195(May), 84–98. False-colour photography: a novel digital
https://doi.org/10.1111/j.1749- approach to visualize the bee view of
6632.2010.05450.x flowers. Journal of Pollination Ecology,
23(12), 102–118.
Lemke, I. H., Kolb, A., Graae, B. J., De Frenne, https://doi.org/https://doi.org/10.26786/1920-
P., Acharya, K. P., Blandino, C., Brunet, J., 7603(2018)11
Chabrerie, O., Cousins, S. A. O., Decocq, G.,
Heinken, T., Hermy, M., Liira, J., Schmucki, Maechler, M., Rousseeuw, P., Struyf, A., Hubert,
R., Shevtsova, A., Verheyen, K., & M., & Hornik, K. (2017). cluster: Cluster
Diekmann, M. (2015). Patterns of phenotypic Analysis Basics and Extensions (R package
trait variation in two temperate forest herbs version 2.0.6.).
along a broad climatic gradient. Plant McCarthy, M. C., & Enquist, B. J. (2007).
Ecology, 216(11), 1523–1536. Consistency between an allometric approach
https://doi.org/10.1007/s11258-015-0534-0 and optimal partitioning theory in global
Lienert, J. (2004). Habitat fragmentation effects of patterns of plant biomass allocation.
fitness of plant populations - A review. Functional Ecology, 21(4), 713–720.
Journal for Nature Conservation, 12(1), 53– https://doi.org/10.1111/j.1365-
72. https://doi.org/10.1016/j.jnc.2003.07.002 2435.2007.01276.x

Liu, C., He, N., Zhang, J., Li, Y., Wang, Q., Sack, Meeus, S., Van den Bulcke, J., & wyffels, F.
L., & Yu, G. (2018). Variation of stomatal (2020). From leaf to label: A robust
traits from cold temperate to tropical forests automated workflow for stomata detection.
and association with water use efficiency. Ecology and Evolution, 10(17), 9178–9191.
Functional Ecology, 32(1), 20–28. https://doi.org/10.1002/ece3.6571
https://doi.org/10.1111/1365-2435.12973 Memmott, J., Craze, P. G., Waser, N. M., & Price,
Liu, W., Zheng, L., & Qi, D. (2020). Variation in M. V. (2007). Global warming and the
leaf traits at different altitudes reflects the disruption of plant-pollinator interactions.
adaptive strategy of plants to environmental Ecology Letters, 10(8), 710–717.
changes. Ecology and Evolution, 10(15), https://doi.org/10.1111/j.1461-
8166–8175. 0248.2007.01061.x
https://doi.org/10.1002/ece3.6519 Mitchell, R. J., Ashman, T., The, S., Phytologist,
Lortie, C. J., & Hierro, J. L. (2022). A synthesis of N., Leake, J. R., Cameron, D. D., & Beerling,
local adaptation to climate through reciprocal D. J. (2008). Predicting Evolutionary
common gardens. Journal of Ecology, Consequences of Pollinator Declines : The
Long and Short of Floral Evolution. New

22
 Phytologist, 177, 576–579. Stevens, M. H. H., … Wagner, H. (2013).
vegan: Community Ecology Package (V. 2.5-
Morison, J. I. L., & Morecroft, M. D. (2008). Plant 7). https://cran.r-project.org/package=vegan
growth and climate change. John Wiley &
Sons. Olivieri, I., Michalakis, Y., & Gouyon, P. H. (1995).
Metapopulation genetics and the evolution of
Murrell, P. (2019). hexView: Viewing Binary Files. dispersal. American Naturalist, 146(2), 202–
https://cran.r-project.org/package=hexView 228. https://doi.org/10.1086/285795
Murtagh, F., & Legendre, P. (2014). Ward’s Ooms, J. (2021). magick: Advanced Graphics and
Hierarchical Agglomerative Clustering Image-Processing in R. https://cran.r-
Method: Which Algorithms Implement Ward’s project.org/package=magick
Criterion? Journal of Classification, 31(3),
274–295. https://doi.org/10.1007/s00357- Opedal, Ø. H. (2019). The evolvability of animal-
014-9161-z pollinated flowers: towards predicting
adaptation to novel pollinator communities.
Naaf, T., Feigs, J. T., Huang, S., Brunet, J., New Phytologist, 221(2), 1128–1135.
Cousins, S. A. O., Decocq, G., De Frenne, https://doi.org/10.1111/nph.15403
P., Diekmann, M., Govaert, S., Hedwall, P.
O., Helsen, K., Lenoir, J., Liira, J., Opedal, Ø. H., Bolstad, G. H., Hansen, T. F.,
Meeussen, C., Plue, J., Poli, P., Spicher, F., Armbruster, W. S., & Pélabon, C. (2017).
Vangansbeke, P., Vanneste, T., … Kramp, The evolvability of herkogamy: Quantifying
K. (2021). Sensitivity to habitat fragmentation the evolutionary potential of a composite
across European landscapes in three trait. Evolution, 71(6), 1572–1586.
temperate forest herbs. Landscape Ecology, https://doi.org/10.1111/evo.13258
36(10), 2831–2848.
https://doi.org/10.1007/s10980-021-01292-w Pickup, M., Field, D. L., Rowell, D. M., & Young,
A. G. (2012). Predicting local adaptation in
Naaf, T., Feigs, J. T., Huang, S., Brunet, J., fragmented plant populations: Implications
Cousins, S. A. O., Decocq, G., De Frenne, for restoration genetics. Evolutionary
P., Diekmann, M., Govaert, S., Hedwall, P. Applications, 5(8), 913–924.
O., Lenoir, J., Liira, J., Meeussen, C., Plue, https://doi.org/10.1111/j.1752-
J., Vangansbeke, P., Vanneste, T., 4571.2012.00284.x
Verheyen, K., Holzhauer, S. I. J., & Kramp,
K. (2021). Context matters: the landscape Poorter, H., Niinemets, Ü., Poorter, L., Wright, I.
matrix determines the population genetic J., & Villar, R. (2009). Causes and
structure of temperate forest herbs across consequences of variation in leaf mass per
Europe. Landscape Ecology, 7. area (LMA): A meta-analysis. In New
https://doi.org/10.1007/s10980-021-01376-7 Phytologist (Vol. 182, Issue 3, pp. 565–588).
https://doi.org/10.1111/j.1469-
Nam, B. E., & Kim, J. G. (2020). Flowering 8137.2009.02830.x
season of vernal herbs is shortened at
elevated temperatures with reduced Poorter, H., Niklas, K. J., Reich, P. B., Oleksyn,
precipitation in early spring. Scientific J., Poot, P., & Mommer, L. (2012). Biomass
Reports, 10(1), 1–10. allocation to leaves, stems and roots: Meta-
https://doi.org/10.1038/s41598-020-74566-z analyses of interspecific variation and
environmental control. In New Phytologist
Oksanen, J., Blanchet, F. G., Kindt, R., Legendre, (Vol. 193, Issue 1, pp. 30–50).
P., Minchin, P. R., O’Hara, R. B., Simpson, https://doi.org/10.1111/j.1469-
G. L., Solymos, P., Stevens, M. H. H., 8137.2011.03952.x
Wagner, H., Friendly, M., Kindt, R.,
Legendre, P., McGlinn, D., Minchin, P. R., Porcher, E., & Lande, R. (2005). The evolution of
O’Hara, R. B., Simpson, G. L., Solymos, P., self-fertilization and inbreeding depression
under pollen discounting and pollen

23
 limitation. Journal of Evolutionary Biology, Skov, F., & Svenning, J. C. (2004). Potential
18(3), 497–508. impact of climatic change on the distribution
https://doi.org/10.1111/j.1420- of forest herbs in Europe. Ecography, 27(3),
9101.2005.00905.x 366–380. https://doi.org/10.1111/j.0906-
7590.2004.03823.x
Postma, F. M., & Ågren, J. (2022). Effects of
primary seed dormancy on lifetime fitness of Strauss, S. Y., & Whittall, J. B. (2006). Non-
Arabidopsis thaliana in the field. Annals of pollinator agents of selection on floral traits.
Botany, 129(7), 795–808. In Ecology and evolution of flowers (Vol. 208,
https://doi.org/10.1093/aob/mcac010 Issue January 2006, pp. 120–138).
Qi, Y., Wei, W., Chen, C., & Chen, L. (2019). Svenning, J.-C., Normand, S., & Skov, F. (2008).
Plant root-shoot biomass allocation over Postglacial Dispersal Limitation of
diverse biomes: A global synthesis. Global Widespread Forest Plant Species in Nemoral
Ecology and Conservation, 18(18), e00606. Europe. Ecography, 31(3), 316–326.
https://doi.org/10.1016/j.gecco.2019.e00606 https://doi.org/10.1111/j.2008.0906-
7590.05206.x
Raiche, G., & Magis, D. (2020). nFactors: Parallel
Analysis and Other Non Graphical Solutions Taylor, K., Woodell, S. R. J. J., Taylort, K., &
to the Cattell Scree Test. https://cran.r- Woodellt, S. R. J. (2008). Biological Flora of
project.org/package=nFactors the British Isles: Primula elatior (L.) Hill.
Journal of Ecology, 96(5), 1098–1116.
Revelle, W. (2022). psych: Procedures for https://doi.org/10.1111/j.1365-
Psychological, Psychometric, and 2745.2008.01418.x
Personality Research. https://cran.r-
project.org/package=psych Uzelac, B., Stojičić, D., Budimir, S., Ramawat, K.
G., Ekiert, H. M., & Goyal, S. (2012).
Richardson, A. D., Bailey, A. S., Denny, E. G., Glandular Trichomes on the Leaves of
Martin, C. W., & O’Keefe, J. (2006). Nicotiana tabacum: Morphology,
Phenology of a northern hardwood forest Developmental Ultrastructure, and
canopy. Global Change Biology, 12(7), Secondary Metabolites. In K. G. Ramawat,
1174–1188. https://doi.org/10.1111/j.1365- H. M. Ekiert, & S. Goyal (Eds.), Plant Cell
2486.2006.01164.x and Tissue Differentiation and Secondary
Rosbakh, S., Römermann, C., & Poschlod, P. Metabolites: Vol. 23(2). Springer.
(2015). Specific leaf area correlates with Van Daele, F., Honnay, O., & De Kort, H. (2021).
temperature: new evidence of trait variation The role of dispersal limitation and
at the population, species and community reforestation in shaping the distributional shift
levels. Alpine Botany, 125(2), 79–86. of a forest herb under climate change.
https://doi.org/10.1007/s00035-015-0150-6 Diversity and Distributions, 27(9), 1775–
Schou, O. (1983). The distyly in Primula elatior 1791. https://doi.org/10.1111/ddi.13367
(L.) Hill (Primulaceae), with a study of Van Daele, F., Honnay, O., & De Kort, H. (2022).
flowering phenology and pollen flow. Genomic analyses point to a low evolutionary
Botanical Journal of the Linnean Society, potential of prospective source populations
86(3), 261–274. for assisted migration in a forest herb.
https://doi.org/10.1111/j.1095- Evolutionary Applications, 2022(00), 1–16.
8339.1983.tb00973.x https://doi.org/10.1111/eva.13485
Schumaker, N. H. (1996). Using landscape Van Rossum, F. (2008). Conservation of long-
indices to predict habitat connectivity. lived perennial forest herbs in an urban
Ecology, 77(4), 1210–1225. context: Primula elatior as study case.
https://doi.org/10.2307/2265590 Conservation Genetics, 9(1), 119–128.

24
 https://doi.org/10.1007/s10592-007-9314-2 Phytogeographical evidence for post-glacial
dispersal limitation of European beech forest
Vandepitte, K., Jacquemyn, H., Roldán-Ruiz, I., & species. Ecography, 32(6), 1011–1018.
Honnay, O. (2007). Landscape genetics of https://doi.org/10.1111/j.1600-
the self-compatible forest herb Geum 0587.2009.05957.x
urbanum: effects of habitat age,
fragmentation and local environment. Xu, Z., & Zhou, G. (2008). Responses of leaf
Molecular Ecology, 16, 4171–4179. stomatal density to water status and its
https://doi.org/10.1111/j.1365- relationship with photosynthesis in a grass.
294X.2007.03473.x Journal of Experimental Botany, 59(12),
3317–3325.
Verheyen, K., Honnay, O., Motzkin, G., Hermy, https://doi.org/10.1093/jxb/ern185
M., & Foster, D. R. (2003). Response of
forest plant species to land-use change: A Zhang, D. (2021). rsq: R-Squared and Related
life-history trait-based approach. Journal of Measures (V. 2.2). https://cran.r-
Ecology, 91(4), 563–577. project.org/package=rsq
https://doi.org/10.1046/j.1365-
2745.2003.00789.x Zhang, W., Hu, Y. F., He, X., Zhou, W., & Shao, J.
W. (2021). Evolution of Autonomous Selfing
Vittoz, P., & Engler, R. (2007). Seed dispersal in Marginal Habitats: Spatiotemporal
distances: A typology based on dispersal Variation in the Floral Traits of the Distylous
modes and plant traits. Botanica Helvetica, Primula wannanensis. Frontiers in Plant
117(2), 109–124. Science, 12(December), 1–15.
https://doi.org/10.1007/s00035-007-0797-8 https://doi.org/10.3389/fpls.2021.781281
Weathers, K. C., Cadenasso, M. L., & Pickett, S.
T. A. (2001). Forest edges as nutrient and
pollutant concentrators: Potential synergisms
between fragmentation, forest canopies, and
the atmosphere. Conservation Biology,
15(6), 1506–1514.
https://doi.org/10.1046/j.1523-
1739.2001.01090.x
Weber, A., & Kolb, A. (2011). Evolutionary
consequences of habitat fragmentation:
Population size and density affect selection
on inflorescence size in a perennial herb.
Evolutionary Ecology, 25(2), 417–428.
https://doi.org/10.1007/s10682-010-9430-1
Whale, D. M. (1983). The Response of Primula
Species to Soil Waterlogging and Soil
Drought. International Association for
Ecology, 58(2), 272–277.
Wickham, H. (2016). ggplot2: Elegant Graphics
for Data Analysis. Springer-Verlag New York.
https://ggplot2.tidyverse.org
Wickham, H. (2021). tidyr: Tidy Messy Data.
https://cran.r-project.org/package=tidyr
Willner, W., Di Pietro, R., & Bergmeier, E. (2009).

25

View publication stats