Leaf‐level resistance to frost, drought and heat covaries across European temperate tree seedlings

Most trees die as seedlings, with harsh environmental conditions, such as early‐spring frosts and summer heat waves, being important drivers of early mortality. However, it remains unclear whether tolerance to different environmental extremes (e.g. frost vs. heat) trades‐off, or covaries synergistically and how stress tolerances relate to growth rates and life history strategies. Given the likely role of extreme environmental conditions as environmental filters, the ability to tolerate different stressors at the seedling stage could shape the occurrence and composition of present and future forests. We explored the relationships between different leaf‐level stress tolerances, functional traits and geographic distributions across 22 species of temperate European tree seedlings. We measured indicators of tolerance to frost, drought and heat and related these values to growth rates and to important functional traits (e.g. leaf mass per area and stem‐specific density). Finally, we explored the links between measured seedling stress tolerances and climatic niche limits inferred from adult trees' distributions. We found that seedlings of most species were either moderately tolerant to all three stressors, or susceptible to all of them. Moreover, higher stress tolerances were associated with traits describing slower growth and lower competitive ability. However, seedling tolerances to climatic factors were unrelated to the environmental limits of their adult geographic distributions. Synthesis. Our results suggest that temperate tree seedlings might not experience trade‐offs when facing an increase in multiple extreme climate stressors, but may experience trade‐offs related to growth rate and competitive ability in the establishment phase. The lack of correlation between leaf‐level stress tolerances and the environmental limits of adult geographic distributions suggests that predicting species' current or future geographic distributions in Europe will require a more nuanced understanding of how climatic tolerances at juvenile and adult stages influence range limits. A better understanding of the interaction between survival in extreme climate, leaf‐level stress tolerances of seedlings and the factors driving species distributions is needed to understand future forest responses to climate change.


| INTRODUC TI ON
Most trees die as seedlings (Harper, 1977).In temperate and boreal forests, a large proportion of this mortality is due to extreme environmental conditions, for example, frost and drought (Stein & Kimberling, 2003;Will et al., 2013).As climate and the frequency of extreme environmental conditions change (IPCC, 2021), currently forested areas might become unsuitable for tree seedlings (Davis et al., 2019).For example, by the end of the century, climate in Central Europe will likely resemble climates currently experienced in Southern Europe (Beniston et al., 2007).In addition, extreme weather events are becoming more common, with more intense and frequent heatwaves, droughts and flooding (Balting et al., 2021;IPCC, 2021;Perkins-Kirkpatrick & Lewis, 2020) and increased risks of early-spring frost in certain areas (Zohner et al., 2020).Tree species may not be able to recruit in the new climatic conditions (Dyderski et al., 2018;Mauri et al., 2022), ultimately reducing forest cover.However, areas currently unsuitable for trees could turn into forests, if tree seedlings are able to recruit in new areas (e.g.above the altitudinal tree line).Consequently, seedlings' ability to tolerate environmental stress will likely influence where forests occur in future (Pozner et al., 2022).Even in areas that are currently forested and will remain so in future, differences among species in their tolerance to environmental stress could influence community turnover.Because changes in forest cover and composition can have large impacts on ecosystem carbon balance and biodiversity (Knapp et al., 2008;Koch et al., 2015), understanding these links is important.
Although there are studies providing insight into the role of seedling stress tolerance and survival on forest occurrence, key knowledge gaps remain.First, many studies have focussed on individual stress tolerances, but few quantify sensitivity to multiple climatic factors, especially across multiple species (but see e.g.Apgaua et al., 2019;Niinemets & Valladares, 2006).This means our understanding of how species will respond to multiple climatechange-related stress tolerances remains incomplete.Although some studies demonstrate that seedlings' tolerance to one stressor can hamper their ability to tolerate another (Eränen et al., 2009), others have found no evidence for trade-offs between stress tolerances (Sack, 2004;Sánchez-Gómez et al., 2006).Studies on trees show equally variable relationships between stress tolerances.
For example, drought and waterlogging tolerances correlate negatively across Northern woody species, as do drought and shade (Niinemets & Valladares, 2006;Puglielli et al., 2021).By contrast, among subtropical tree species, drought and frost tolerance correlate positively across conifers, but not across deciduous species (di Francescantonio et al., 2020).Moreover, results from studies on trees are unlikely to apply to seedlings, because many of the physiological processes and structural traits relevant for stress tolerance change with age (Niinemets, 2010;Wieser et al., 2003).Furthermore, how differences in environmental tolerances translate to growth or survival differences, and ultimately, range shifts and compositional turnover, is also poorly understood.
A second knowledge gap is how species tolerances to different climatic stressors relate to competitive ability at seedling stages.
Based on the trade-off between stress tolerance and competitive ability (Grime, 1977), slow-growing species with resourceconservative traits are expected to be more stress-tolerant but also more sensitive to competition compared with fast-growing species with resource-acquisitive traits.Some studies show support for this trade-off (Hallik et al., 2009;Keep et al., 2021), but others do not (Fernández & Reynolds, 2000;Jung et al., 2020).For example, across Northern hemisphere woody taxa, drought tolerance has been found to be associated with resource-conservative traits, whereas shade tolerance can be achieved with different sets of traits along the leaf economic spectrum (Hallik et al., 2009).Thus, it remains unclear how traits describing the fast-slow life history axis relate to stress tolerances at the seedling stage, especially when stress tolerances themselves may be negatively correlated.
Finally, it is unclear whether species traits at the seedling stage, especially those relevant to environmental tolerances, can help us understand forest responses to climate change.One test of this is understanding how seedling traits are related to the species' current geographic distributions.Studies on the relationship between traits and distributions often concentrate on adult life stages and provide a mixed picture.On the one hand, leaf turgor loss point ('TLP') of adult trees has been linked to water availability within and across biomes (Bartlett, Scoffoni, & Sack, 2012), cold tolerance of North American broadleaf trees to minimum temperatures (Hawkins et al., 2014) and the trade-off between shade and drought tolerance to the distribution of conifer trees in North America (Rueda et al., 2017).On the other hand, frost hardiness of European trees is not related to their macroclimatic niches (Hofmann & Bruelheide, 2015).If forest composition is the result of the seedlings that have survived until adulthood in a given location (HilleRisLambers et al., 2012), and survival is linked to seedling's traits and combined stress tolerances, there should be a clear link between seedling stress tolerance and adult occurrence in climate space.Nevertheless, complicating the matter, survival in extreme climate, leaf-level stress tolerances of seedlings and the factors driving species distributions is needed to understand future forest responses to climate change.

K E Y W O R D S
drought, frost, heat, seedling, stress, temperate forest, tolerance the distributions of most European tree species are thought to not be in equilibrium with their optimal climate due to slow post-glacial recovery and geographic dispersal barriers (Svenning & Skov, 2007).
It is thus unclear whether the physiology of species' stress tolerance, especially at the seedling stage, is linked to the species' distributions in Europe.
We address these issues by exploring how tree seedling traits and stress tolerances, particularly those relevant to climatic vulnerability, are correlated across species and whether they are predictive of the climatic limits of their distributions.In a greenhouse experiment, we measured leaf-level tolerances to three important abiotic stressors (frost, drought and heat) across seedlings of 22 widely distributed temperate European tree species, together with functional traits reflecting species' competitive abilities (Table 1).Such a large set of traits and tolerances are rarely all measured across many species.The frequencies of the three climatic stresses we focussed on (heat, drought and frost) are changing in the regions where the species currently occur (Balting et al., 2021;IPCC, 2021;Perkins-Kirkpatrick & Lewis, 2020;Zohner et al., 2020), and we expected leaf-level measurements of tolerances to influence survival of those species when facing these climate extremes.Finally, we compared the leaf-level stress tolerances of seedlings to the environmental conditions at the 'edges' of these species' potential adult distributions in climate space.In all, we used our data to ask the following specific questions: 1. How do seedlings of different species differ in their leaf-level tolerance to frost, drought and heat?Do these tolerances covary in positive (synergy) or negative (trade-off) ways?As our experiment covered a large range of species with different environmental preferences, we expected to uncover large variation in the three different stress tolerances.We also expected these to covary so that species would express either low or high stress tolerance, but to different stressors.
2. Are seedling stress tolerances predictably related to functional traits describing the fast-slow life history continuum?We expected stress tolerance to be negatively related to traits related to fast growth or high competitive ability (Grime, 1977).

Do seedling stress tolerances to climatic variables predict tree
species distributions in climate space?As climatic stress can shape forest composition by influencing which seedlings survive until adulthood (HilleRisLambers et al., 2012), we expected seedling stress tolerance to be related to adult tree distributions.

| Experimental set-up
We investigated seedling responses of 22 tree species common in temperate European forests.Species included 15 deciduous broadleaves, six evergreen conifers and one deciduous conifer (Figure 1; Table S1).Seeds were collected by the Swiss Federal Institute for Forest, Snow and Landscape Research (WSL) from the locations listed in Table S2, between 2015 and 2020.Seeds were stratified according to WSL guidelines (Burkart, 2018).The germinating seeds (25-30 individuals per species) were planted to 1-L pots 1 April-10 May 2021 in the greenhouse in the ETH campus Hönggerberg (Zürich, Switzerland).In addition, 24 seedling per species were planted in smaller 0.4-L pots for the frost tolerance measurements.
The plants grew in ambient temperature and light conditions and were watered every 2-3 days.The experiment lasted until 31 August 2021.The location of each seedling in the greenhouse was randomized and shuffled once during the experiment.For a more detailed description of all the methods, see Supporting information file S1.

| Measuring frost tolerance
We chose to measure seedling frost tolerance, because early-spring frost can have detrimental effects on many plant species (Hufkens et al., 2012).Frost risk has also been found to play a role in setting upper elevational and latitudinal range limits of some species (Bucher & Rosbakh, 2020;Segovia et al., 2020;but see Hofmann & Bruelheide, 2015).We assessed frost tolerance with the electrolyte leakage method (Pérez-Harguindeguy et al., 2013).Because seedlings are most vulnerable to frost early in the spring and the developmental stage of the seedling can influence their frost tolerance, we carried out the frost treatment once each seedling had unfolded the first two true leaves or the first whorl of true needles (4-13 seedlings per species, see Table S3).Seedlings were subjected to one of the two frost treatments: −4°C (a common temperature for late spring frost in Swiss lowlands, Vitasse & Rebetez, 2018) or −7°C, used to simulate a more extreme frost event (https:// clima te.coper nicus.eu/ esotc/ 2021/ late-sprin g-frost ).Seedlings were placed in a cooling incubator (VWR INCU-Line Premium, Avantor, Pennsylvania, USA) where temperature dropped at a rate of 2°C/h (Steffen et al., 1989) from +4°C to −4°C or to −7°C.Afterwards, two leaf discs cut from each seedling were refrigerated in deionized water for 24 h.Conductivity was measured with a conductivity meter ('LAQUAtwin-EC-22', HORIBA, Japan).Samples were then boiled for 15 min (to fully damage all cells), allowed to cool down, and measured again.Frost resistance was determined as the difference in percentage of damaged cells ('PEL', conductivity before boiling / after boiling) between the treatment and the control (no frost treatment) samples 100 − (PEL treatment − PEL control ) (Pérez-Harguindeguy et al., 2013).

| Measuring drought tolerance
We measured seedling drought tolerance, because drought is often related to seedling mortality (Lopez & Kursar, 2007), and likely to increase in frequency and intensity in the future (Balting et al., 2021).
Leaf-level drought tolerance was measured as the water potential TA B L E 1 Summary of the traits and stress tolerances measured.shows turgor loss point (-TLP) as a measure of drought tolerance, and panels (c) shows a metric describing heat tolerance (T 50 ).Points are mean values per species, error bars are 95% confidence intervals (so that non-overlapping error bars are significantly different from each other at 5% probability).For full descriptions of the stress tolerances, see Table 1 in the main text.For numerical values for all tolerance metrics, see Table S1.

Description
at which turgor pressure is lost (Banks & Hirons, 2019;Bartlett, Scoffoni, Ardy, et al., 2012).This measurement (turgor loss point at wilting, 'TLP') has been shown to correlate with water availability within and across biomes (Bartlett, Scoffoni, & Sack, 2012) and to plant survival under drought conditions (Álvarez-Cansino et al., 2022).There is some uncertainty on how well TLP corresponds to drought tolerance among co-occurring species (see e.g.Farrell et al., 2017), but as our selection of species occurred across a relatively large range of climate and elevation, we considered TLP to be a suitable metric for separating species with drought-tolerant leaves from those with drought-sensitive leaves.We measured drought tolerance for 3-11 seedlings per species (Table S3) by using extracted sap (Barzilai et al., 2021;Callister et al., 2006) and a vapour-pressure osmometer (Vapro 5600, ELITechGroup, Puteaux, France) to determine the concentration of osmotically active compounds in the plant cells at full hydration.We calculated water potential (π o ) and TLP (π tlp ) based on the measured osmotic concentration (Banks & Hirons, 2019;Bartlett, Scoffoni, Ardy, et al., 2012).

| Measuring heat tolerance
We measured leaf-level heat tolerance of seedlings, because the frequency of heatwaves is increasing in Europe (Beniston et al., 2007).
Furthermore, upper thermal limits are evolutionarily conserved across several taxa (Araújo et al., 2013).Thus, extreme heat could limit species distributions in future.Tolerance to heat stress was measured as a reduction in the performance of photosystem II after exposure to heat stress following Perez and Feeley (2020) and Krause et al. (2010).This 'photosynthetic heat tolerance' has been hypothesized to be an adaptation to withstand extreme leaf temperatures (Perez & Feeley, 2020;Zhu et al., 2018), even though its relationship to extreme air temperatures and whole plant heat tolerance is still unclear.We measured the ratio between variable chlorophyll fluorescence to maximum fluorescence (Fv/Fm) from dark-adapted leaf discs after 15-min exposure to the following temperatures: room temperature (control), 38,40,42,44,46,48,50,52, 54 and 60°C.Leaf discs were cut from leaves from 2 to 6 seedlings per species (with one individual providing discs for all 11 treatments, Table S3).Leaf discs were enclosed in water-proof plastic bags and submerged into a waterbath set to one of the treatment temperatures.Afterwards, the leaf discs were dark-adapted for 20 min, and Fv/Fm was measured using Multispeq (PhotosynQ, Michigan, USA).
We fitted a logistic nonlinear model to the data using the modelling approach of Perez and Feeley (2020), from which we extracted three parameters: (1) temperature at which Fv/Fm starts to decline (i.e. slope of the response is 15% of the maximum slope 'T crit '), (2) temperature at which Fv/Fm has dropped 50% of the maximum ('T 50 ') and (3) temperature at which Fv/Fm has dropped 95% ('T 95 ').For responses of all species, see Figure S1.
To verify that measurements on leaf discs reflect heat tolerance of whole leaves or seedlings, we performed two additional experiments.Briefly, for a small subset of the species, we either heated whole leaves while still attached to the plant, or placed whole seedlings into a heating incubator.Fv/Fm was measured after 15-min exposure to 45°C.See Supplementary information file S2.

| Measuring growth rate and functional traits
Plant functional traits can serve as proxies of species' general life history, form and function (Díaz et al., 2016;Wright et al., 2004) and together with growth rate, relate to species' competitive ability and stress tolerance (Grime, 1977).We measured relative growth rate ('RGR'), above-to below-ground biomass ratio ('root:shoot'), leaf area ratio ('LAR') and three traits related to species position on the fast-slow life history axis: leaf mass per area ('LMA'), leaf dry matter content ('LDMC') and stem-specific density ('SSD', Table S4).In June 2021, we chose 10 individuals per species (when available, see Table S3), and divided them into five pairs of similar size.One seedling of each pair was harvested between 23 June and 19 July, the other seedling between 31 August and 2 September.We first washed the root system, then chose one fully expanded and healthy leaf or needle and weighed and scanned them immediately after being detached from the stem.We then measured the length of the stem, and its diameter at the base and at the tip.The complete seedlings were dried in 70°C for 3-4 days, to measure their dry weight.Leaf area was determined from scanned images using ImageJ (Abramoff et al., 2004).To estimate RGR, we calculated the difference in total biomass per species between June and September, divided by the number of days in between measurements and by seedlings' initial size.For the other traits, we calculated average trait value by combining the early-and late season measurements.

| Trade-offs between stress tolerances, and between-stress tolerances and functional traits
To test how tolerances to different stressors trade-off with each other, we inspected pairwise stress tolerance combinations using Deming regressions.We chose a Deming regression approach, because both response and explanatory variables are associated measurement errors (variation in stress tolerances between individuals within a species) and were not measured on the same individuals.
To further investigate how different species were positioned in the stress tolerance space, we performed a phylogenetically corrected principal component analysis (PCA; Revell, 2009) with all the stress tolerance metrics (T crit , T 50 , T 95 , TLP, PEL at −4°C and PEL at −7°C).
To test how stress tolerances relate to species' growth rate and functional traits, we performed a phylogenetically corrected PCA with the measured functional traits (LMA, SSD, LDMC, root:shoot, RGR, LAR).To obtain the phylogenetic relationship of study species, we pruned the DaPhnE 0.1 supertree of the European vascular plant species (Durka & Michalski, 2012).We used the first two PCA axes (Figure S2b) to build linear models, with stress tolerance metric as a function of the two trait axes (i.e.stress ~axis1 + axis2).Significance of the fixed effects were tested using likelihood ratio tests.

| Relationships between species' stress tolerance and climatic niche
To compare leaf-level tolerance values measured in the greenhouse with those estimated from species' distributions, we modelled each species' potential distribution using species distribution modelling (SDMs).We used this potential distribution as a proxy of the species' fundamental niche, and then calculated the fundamental range edge based on long-term climatic values most closely related to the experimentally measured tolerance variables.For presence data, we used species' potential presence probability maps (Mauri et al., 2022), rather than records of observations, to ensure consistency in record data and modelling choices across species and space.
We used the ensembled potential current distribution for each species as the closest metric to the species potential climatic niche, aiming to reduce the influence of unnatural disturbances or land use in these models.The potential distribution maps were available for 20 of our 22 species (excluding Sorbus torminalis, Pinus mugo).We used multi-model ensembles for each species, which combine model estimates from different modelling approaches, to improve overall model performance and reduce the risk of prediction bias by a particular algorithm (Robinson et al., 2017).
We used each species' potential distributions to obtain their tolerance values at the edge of the species distribution using three widely employed climate extreme variables from the WorldClim2.0 climate database (Fick & Hijmans, 2017).Namely, we selected were found to be highly correlated, particularly for the first months (March, April) (Figure S3).Finally, to look for single events, we calculated annual last spring frost date values (LSF) as the last calendar date of the year with high risk of frost occurrence (Ma et al., 2019) and related it to the bioclimatic variables and the experimentally obtained physiological tolerances (Figure S4).Note that we use the term 'frost tolerance' to refer to the tolerance metric measured via electrolyte leakage in the greenhouse, and 'cold tolerance' to refer to the cold edge of the species' fundamental niche.
We also found that the temperature which was critical for most species when measured as leaf discs (45°C) was less so when measured on whole leaves or seedlings: functioning of the photosystem II (Fv/ Fm) was reduced less in whole leaves and seedlings and recovered after couple of days (Supplementary file S2).For many of the traits and tolerances, we observed distinct differences between conifers and broadleaves, which tended to cluster separately.This was true especially for the first axis of the PCA including functional traits (governed by LAR, LMA and root:shoot), which separated conifers and broadleaves in opposite sides of the trait space (Figure S2b).
We found little evidence for trade-offs in leaf-level adaptation to multiple climatic stressors.Instead, species with high tolerance to frost tended to also have high tolerance to drought and extreme heat (Figure 2; Table S5).The first axis of the PCA including all the stress tolerance metrics explained 61% of the variation and separated species with high tolerance to all stressors from those with low tolerance to all stressors (Figure S2a; Table S6).Despite the lack of trade-offs between tolerances to different stressors, we did observe a trade-off related to how different species tolerate heat stress.The temperature at which leaf photosynthetic capacity started declining (T crit ) correlated negatively with the tolerance to extreme heat (T 95 , Figure S5b).In other words, species that were better able to tolerate an initial increase in temperature had a lower tolerance for extreme heat (Figure S6).The second axis of the PCA built with the stress tolerance metrics (explaining 31% of the variation) separated species along this axis: from high initial resistance to heat (high T crit ) and low tolerance to extreme heat (low T 95 ) to species with opposite heat resistance traits.The same axis separated species with high tolerance to either moderate (−4°C) or extreme (−7°C) frost.

| Higher stress tolerance was related to slow growth and to traits related to low competitive ability
We found that the leaf-level stress tolerance metrics were correlated with the functional trait axes.The first axis of the PCA constructed from functional traits (LMA, LDMC, SSD, root:shoot, LAR and RGR) explained 69% of the variance and separated species that grow quickly and invest in large leaf area at the expense of leaf robustness (i.e.high LAR and RGR, low LMA and LDMC) from species with opposite traits (Figure S2b).Effectively, this axis separated conifers from broadleaves.The second PC axis explained 31% of the variance and separated species with high root:shoot ratio and low SSD and LMA from species with opposite traits.The first PC axis (conifer-broadleaf) correlated positively with tolerance to extreme heat (T 50 and T 95 ) and drought (Figure 3a,b; Table S6) indicating that species with lower investment in leaf area and low growth rate, but with more robust leaves (i.e.conifer species) were more tolerant to extreme heat and drought.The second PC axis (SSD, root:shoot) correlated positively with tolerance to drought (Figure 3c), suggesting that species with high root-to-shoot ratio and low stem density were more tolerant to drought.Heat tolerance was not related to the second PC axis (Figure S7).Frost tolerance was not related to either PC axes (Figure S7).As expected, physiological drought tolerance (i.e. TLP) was associated with morphological traits related to drought tolerance (higher root:shoot, lower LAR, higher LMA and LDMC).

| Seedling stress tolerance was unrelated to modelled climatic range limits
We found little to no relationship between leaf-level stress-tolerance of seedlings measured in the greenhouse and the conditions at the  S5).Blue points are broadleaves, pink triangles are conifers.For simplicity, we show correlations only for one metric of heat and frost tolerance (T 50 and PEL at −7°C).For the remaining pairwise comparisons, see Figure S5.For descriptions of the tolerance metrics, see Table 1.
'edges' of climate space occupied by adult trees (Figure 4; Figures S8- S10).The relationships between leaf-level stress tolerances and climate edges, that is, frost tolerance (PEL at −7°C) with the minimum temperature of the coldest month occupied by trees (Tmin); heat tolerance (T 50 ) with maximum temperatures of the warmest month occupied by trees (Tmax), and turgor loss point (-TLP) with drought indices or precipitation in the warmest part of the year (Pwarm and ldM), showed small correlation values.Furthermore, the hypothesized correlations were weaker than those between seemingly unlinked phenomena (e.g. between frost tolerance and precipitation of the warmest quartier).The absence of strong relationships between stress tolerances and distributional limits in climate space was consistent across a wide range of metrics used to quantify niche-based tolerance to environmental extremes and multiple quantile values to calculate edges (Figure S8).
Interestingly, drought tolerance measured in the greenhouse (-TLP) showed high variation for broadleaves, whereas the dry edge of the climate space occupied by adult trees (Pwarm or IdM) varied only little among broadleaves.By contrast, the opposite was true for conifers, where we found a large spread in conditions on the dry edge of their climatic niche, while their experimentally measured tolerances were very close to each other (Figure S9).The relationships between experimentally measured drought tolerance and niche-based drought tolerance remained non-significant even when broadleaves and conifers were tested independently, indicating that the lack of a significant relationship was not caused by broadleaf/ conifer differences (Figure S10).

| DISCUSS ION
We found considerable variation in leaf-level stress tolerance across 22 species of temperate tree seedlings, especially for drought and frost, but little evidence of trade-offs in adaptation to multiple climatic stressors.In fact, most species were either relatively tolerant to all three stressors, or susceptible to all, suggesting that adaptation to stress tolerance in general is consistent across species.We did find support for a trade-off between leaf-level stress tolerance and traits often related to fast growth or high competitive ability.Finally, we found that seedling stress  S7) correlations are shown.Blue points are broadleaves, pink triangles are conifers.For the PCA, see Figure S2b and for correlation between PCA axes and all the stress tolerance metrics, see Figure S7.LMA refers to leaf mass per area, LDMC to leaf dry matter content LAR to leaf area ratio, RGR to relative growth rate and SSD to specific stem density (Table 1).
tolerance measured in the greenhouse was not related to the edge of the climatic niche of the species.This suggests that occurrencebased models are unlikely to capture regeneration tolerances to extremes across age stages and that physiologically informed demographic models are likely required to capture these responses (see e.g.Ghosh et al., 2016;Shriver et al., 2022).Moreover, information on how leaf-level measurements of stress tolerance correspond to whole-plant survival, and how microclimate might mediate that, are needed to better understand the links between regeneration, climate, and forest occurrence.We discuss these findings below.

| Stress tolerance differed between species and covaried positively
We found that leaf-level stress tolerances covaried across the studied species.This is similar to previous findings on Northern hemisphere woody species: Most species show intermediate tolerance to multiple stressors whereas very few species show extreme tolerance (Puglielli et al., 2021).In general, physiological constraints are considered to prevent poly-tolerance to multiple stressors (Smith & Huston, 1990), as might be the case with the previously documented shade-drought tolerance trade-off (Laanisto & Niinemets, 2015).However, our results suggest this may not be the case for all stressors.Tolerance to different stressors might mutually reinforce each other if these stressors induce common physiological responses (Kreyling et al., 2014;Mayr et al., 2006).For example, drought and frost are related to limited water availability and result in similar damage at the cellular and vascular levels (Siminovitch & Cloutier, 1983).Responses to both of these stressors includes accumulation of solutes and a 'solvation' layer around macromolecules and activation of the same molecular pathways (Yamaguchi-Shinozaki & Shinozaki, 1994), as well as thick, tough leaves with high capacity for water storage.Moreover, all of the three stressors measured (heat, drought, frost) are likely to occur in temperate forests at some point during the growing season (Niinemets, 2010), exerting a clear selective pressure towards poly-adaptation.
For some of the species, the stress tolerance values measured in our experiment differed from what has been previously found for adult trees (Table S8).These discrepancies could arise because tolerance to environmental extremes at the seedling stage differs from that of adult trees, or because our measurements do not capture all aspects of stress tolerance.Importantly, our measurements reflect stress tolerance at the leaf-level, which might not correlate with effects of stress on whole seedling growth and mortality.Species might tolerate climatic extremes for example by resprouting (Pausas et al., 2016), or by adjusting phenology, as is the case with frost tolerance (Zohner et al., 2020).Further validation on whether the leaflevel metrics of stress tolerance correspond to whole plant survival (see e.g.Álvarez-Cansino et al., 2022) in the face of extreme events is needed.
Our metrics for heat tolerance were also limited in the sense that they reflect only leaf-level maximum temperatures, and do not take into account species' ability to thermoregulate (Drake et al., 2020;Rey-Sánchez et al., 2016).Adding to this complexity, our experiments revealed that individual leaves can recover from heat stress, and whole seedlings might tolerate higher temperatures than leaf discs (Supplementary file S2).These results highlight that although our methods might be suitable for uncovering differences between species in leaf-level heat tolerance, they may not reflect actual critical temperature limits for the whole organism (relative to fitness).Especially, species' ability to recover from heat stress could be a key characteristic determining survival in novel, hotter climates.Accurately determining critical heat limits Correlation matrix showing the relationship between directly measured leaf-level stress tolerance (PEL at −7°C, T 50 , and -TLP) and stress tolerances calculated from species climatic niche (Tmin, Tmax, Pwarm, IdM) at their 50th, and 5th or 95th percent quantile.See Table 1 for descriptions of the variables.Positive and negative correlations are highlighted in blue and red respectively.The correlations between stress tolerances (rows) and climatic niche limits (columns) which we expected to be strongest (i.e. between stress tolerance metric and corresponding climatic variable) are highlighted with black boxes.The relationship between the selected climatic variables and other quantiles were highly consistent (Figure S8).
and species' ability to recover from heat stress, and extending experiments to study mortality will be important with the increased frequency and intensity of heat waves in the future (Perkins-Kirkpatrick & Lewis, 2020).Indeed, many of the critical heat limits measured experimentally were exceeded in Europe during summer 2021 (https:// clima te.coper nicus.eu/ esotc/ 2021/ medit erran ean-summe r-extremes).Thus, whether these leaf-level limits are good proxies for actual temperature limits for survival and regeneration should be investigated.

| Higher stress tolerance was related to slow growth and to traits related to low competitive ability
In contrast to the lack of trade-offs when comparing different climatic tolerances, we did find that higher stress tolerance was associated with slower growth rates and with traits related to slow life history strategy (high LDMC, high LMA, Figure 3).Even though relationships between stress tolerance and growth rate can vary, the general expectation is that high stress tolerance trades-off with fast growth rate, typical for competitive or ruderal species (Grime, 1977).
The trade-off exists partly because the structural and physiological adaptations that allow fast growth rate also make plant tissue more vulnerable to environmental extremes (e.g.large and less robust leaves; Hallik et al., 2009).One manifestation of this trade-off is the difference in leaf structure between evergreen conifers and deciduous broadleaf species.Longer leaf lifespan correlates with traits related to higher stress tolerance (Wright et al., 2004).Thus, evergreen and deciduous species fall to opposite ends on the leaf trait spectrum, as seen on the first axis of the PCA including leaf traits (Figure S2b) and in previous studies (Díaz et al., 2016;Hallik et al., 2009).Nevertheless, within both functional groups, higher stress tolerance can be achieved with a different combination of traits (Hallik et al., 2009).Our results support a general trade-off between competitive ability/fast growth rate and general stress tolerance between and across the two functional groups.
Several areas across Europe are likely to become more stressful for tree seedlings in the future due to increased heatwaves, droughts, floods and changes in spring frost risk (Balting et al., 2021;IPCC, 2021;Perkins-Kirkpatrick & Lewis, 2020;Zohner et al., 2020).
In these areas, we expect slow-growing species with higher stress tolerance to become more common.Vice versa, areas becoming less stressful for tree growth (e.g.longer growing season) could see an increase in fast-growing and less stress-tolerant species.
Tree growth rate can influence its lifespan and, consequently, the residency of sequestered carbon by forests (Büntgen et al., 2019;Piovesan & Biondi, 2021).Indeed, faster growing species tend to have lower wood density and die younger (Piovesan & Biondi, 2021).
The increased carbon sequestration by faster growth rate might be reduced or reversed by shorter lifespan and thus shorter carbon residency (Brienen et al., 2020;Büntgen et al., 2019).How changes in climate extremes and consequently forest composition affect future carbon sequestration by forests requires further investigation.

| Seedling stress tolerance was unrelated to modelled climatic range limits
We found that the experimentally measured stress tolerance metrics of tree seedlings did not relate to the edges of the potential climatic distributions of the species across Europe.The common assumption is that species' climatic niche can be modelled across developmental stages using species' geographic distribution.At the macroscale, abiotic conditions and species' stress tolerances are thus expected to determine distributions, especially for widely distributed species (Hawkins et al., 2003;Pither, 2003).In practice, this is not always the case (Rosbakh et al., 2020;Wisz et al., 2013).For example, the Pinus cembra/Pinus sylvestris ecoline in the Alps seems to be set by competition, mechanical damage by large herbivores and habitat preferences of the seed dispersers, creating a distribution pattern along a moisture gradient opposite of the physiological drought tolerances of the two species (Hättenschwiler & Körner, 1995).Furthermore, species distribution models, like those used here, are commonly assumed to represent the suitable conditions for the whole speciesbut are often not estimated using data from all life stages.This is true especially for trees, where SDMs are fit predominantly using presence of adult stages only (but see e.g.Ghosh et al., 2016).It is currently unclear how the climatic niche of seedlings correlates with that of adult trees, and whether this could distort the relationships between seedling stress tolerance, climate and adult occurrence.
This highlights the need to improve our mechanistic and methodological approach for long-lived widespread species, such as trees.
Lastly, it is important to acknowledge that the distribution of trees in Europe has been heavily influenced by human activity for thousands of years (with natural forest currently representing only 3% of the total forest cover, FAO, 2020).Current patterns of tree presence are likely heavily shaped by human activities in non-random ways that may bias most SDM approaches.
Additionally, historic dispersal barriers restrict distributions of many European species, some of which are still in the process of expanding North since the last ice age (Svenning & Skov, 2007).Low dispersal rates and historical dispersal barriers could create mismatches between traits and distributions.Nevertheless, we do not think this creates a problem for our approach of comparing stress tolerances and distributional edges.First, dispersal limitation since the last ice should mainly affect patterns of cold tolerance, where some species may not have yet filled all the cold edges of their ranges.However, the relationship between frost tolerance and the cold edges of the distributions did not stand out from the other tolerance/distribution relationships.Second, we used the fundamental niche in our modelling.This means that the modelled climate space included climatic variation caused by topographic variation (e.g.elevation).Therefore, the cold limit of the species niche was likely included in the model through, for example, elevational tree lines.
To fully disentangle how and whether seedling stress tolerance influence tree distributions in Europe, there are a couple of aspects that would be needed to take into consideration.First, the occurrence of environmental stress at a given location is influenced by microclimate (Schweiger & Beierkuhnlein, 2016).For example, trees dampen the extreme temperatures under the canopy (Renaud & Rebetez, 2009), allowing seedlings to establish in regions where they would not survive if growing exposed (D'Odorico et al., 2013).Thus, macro-climatic conditions might not offer a good estimate of the intensity of environmental stress a seedling experiences.Second, different populations of the species might be locally adapted (Barton et al., 2020;Kreyling et al., 2014), allowing, for example, a population at the range edge to persist in conditions in which the species on average would not.As our seedlings originated from a single location (per species), we were not able to test for within-species differences in stress tolerance between different populations.Nevertheless, previous research shows that at least heat tolerance is conserved across lineages (Araújo et al., 2013) and therefor unlikely to show strong local adaptation.Similarly, previous studies have failed to find local differences in stress tolerance among tree seedlings (Gimeno et al., 2008;Latreille & Pichot, 2017).Our source populations were also situated in the middle of the ranges of the different species, and most likely represent an average tolerance and trait combination for each species.If populations along a climatic gradient are strongly adapted to different conditions, studying many populations across each of the species would provide more insight into the role local adaptation might play.
Whether the fundamental (climatic) niche of adult trees correspond to climatic conditions suitable for seedlings has practical implications for forest restoration.Often, only a small percentage of the planted seedlings survive until adulthood (Howe et al., 2020).
The high seedling mortality can be due to inappropriate selection of species for a specific site (Howe et al., 2020).The selection of species in restoration projects is usually based on modelling of current species distributions and predicted future habitat suitability (Castro et al., 2021).Our study suggests that current distributions of species might not be indicative of the environmental conditions best suited for seedlings.More focus on which abiotic and biotic conditions are limiting seedling survival will likely benefit forest restoration efforts.

| CON CLUS IONS
We measured leaf-level tolerance to heat, drought and frost in seedlings of 22 species of European temperate trees, and related their stress tolerance to functional traits and climatic niches.We found tolerance to the three stressors to covary, general stress tolerance to be related to seedling growth rate and life history strategy, and unlinked from the modelled climatic distributions of the whole species based on standard SDM approaches.In all, our results therefore hint that current modelling approaches do not capture tolerance to environmental stress at the seedling stage.How leaf-level stress tolerance relates to whole plant survival in extreme climates, and how tolerance to stressors at the seedling stage affects the whole species responses to climate extremes remain fundamental knowledge gaps in our understanding of forest responses to climate change.In future, the intensity of especially drought and heat stress are likely to increase (Balting et al., 2021;Perkins-Kirkpatrick & Lewis, 2020), whereas changes in frost risk are less predictable (Zohner et al., 2020).Our results suggest that, even as climate becomes more variable, temperate tree seedlings might not face trade-offs in terms of adapting to future changes in the risk of heat, frost and drought.

S U PP O RTI N G I N FO R M ATI O N
Additional supporting information can be found online in the Supporting Information section at the end of this article.

Table S1:
The studied species and the mean values for the leaf-level stress tolerance metrics.
Table S2: Information on the origin of the seeds used in the greenhouse experiment.
Table S3: Sample sizes and age at the time of frost resistance measurements.
Table S4: Mean values for functional traits per species.
Table S5: Model results from the Deming regressions investigating relationships between different stress tolerances.
Table S6: Loadings of the different traits influencing the two first axes of the two PCAs.
Table S7: Model results from the linear regressions investigating the relationship between stress tolerances and the two PC axes describing species functional traits.
at wilting, i.e. the leaf water potential at which cells become flaccid Drought tolerance: note that we use (−1) × TLP, so that higher values correspond to higher drought tolerance(Bartlett, Scoffoni, & Sack, 2012)    Stress toleranceTcrit °C 'Critical temperature', i.e. at which the function of photosystem II (measured as Fv/Fm) starts declining Heat tolerance.Higher values signify higher tolerance (Perez & Feeley, 2020) Stress tolerance T 50 °C Temperature at which Fv/Fm is 50% of its maximum value Heat tolerance.Higher values signify higher tolerance (Perez & Feeley, 2020) Stress tolerance T 95 °C Temperature at which Fv/Fm has declined 95% of its maximum value Heat tolerance.Higher values signify higher tolerance (Perez & Feeley, 2020percentage of damaged cells between treatment and control samples after exposure to −4°C Frost tolerance.Higher values signify higher tolerance (Pérez-Harguindeguy et alpercentage of damaged cells between treatment and control samples after exposure to −7°C Frost tolerance.Higher values signify higher tolerance (Pérez-Harguindeguy et al., 2013) Stress tolerance LMA g/m 2 Leaf mass per area Higher values relate to tougher leaves and longer leaf lifespan.Higher LMA gives better protection from physical hazards (Poorter et al.to tougher leaves and longer leaf lifespan.LDMC is often closely related to LMA, and higher values often mean better protected leaves (Pérez-Harguindeguy et al.to denser and more robust stems.SSD is important for stability, defence, architecture, hydraulics, C gain and growth potential of plants (Pérez-Harguindeguy et al., 2013) Functional trait root:shoot Ratio 'Root to shoot ratio', i.e. the ratio of below-ground to above-ground biomass Higher values correspond to higher investment in below vs above-ground structures and to a higher ability to forage resources (water, nutrients) below-ground (Pérez-Harguindeguy et al.area ratio, i.e. the total leaf area divided with the dry mass of the whole plant Higher values correspond to higher investment in carbon capture through leaves and to lower drought resistance Functional trait RGR mg/day Relative growth rate Higher values relate to faster growth rate and to more competitive/ruderal strategy (Grime & Hunt, 1975) , calculated from annual values of temperature (T) and precipitation (P) as P/(T + 10) Drought tolerance.Lower values correspond to drier conditions and higher tolerance Climatic variable Note: The 'Type of trait' column separates measurements of leaf-level stress tolerance directly ('stress tolerance'), measurements of functional traits that describe life history more generally ('functional traits'), and measurements on based on adult tree distributions from species distribution databases ('climatic variables').Stress tolerances and functional traits were measured from greenhouse-grown seedlings.F I G U R E 1 Differences in the three main stress tolerance metrics between species.Panel (a) shows tolerance to −7°C frost, panel (b) Maximum temperature of the warmest month (bio5, hereafter Tmax) as a proxy for heat tolerance, Minimum temperature of the coldest month (bio6, Tmin) for cold tolerance and Precipitation of the warmest quarter (bio18, Pwarm) for drought tolerance.Additionally, we calculated the DeMartonne Aridity Index (IdM) from annual values of precipitation (P) and temperature (T) as (IdM = P/(T + 10); de Martonne, 1926), a widely used measurement of drought influence on tree performance (e.g.Bhuyan et al., 2017 and references therein).The cold, hot and dry edges of the distribution were calculated as the 90, 95 and 99th percentiles for Tmax; and the 10th, 5th and 1st percentiles for Tmin and Pwarm distributions.The percentiles 5th and 95th were used for further calculations, followingZhu et al. (2012).To check the validity of Tmin as a proxy for cold temperatures during spring, we compared the response patterns using Tmin with the minimum temperatures of early spring months (March, April, May), which are the key months for frost risk for seedlings across Europe.These variables

F
Relationships between leaf-level frost, heat and drought tolerance.Panel (a)  shows the relationship between frost and heat tolerance, (b) between drought and heat tolerance and (c) between frost and drought tolerance.Points are mean values per species, error bars ±1 SEM.Dashes lines are predictions from the Deming regressions.Only correlations in which the slope with the associated 95% confidence intervals does not cross 0 (i.e.'significant') are shown (Table

F
Correlations between different metrics for leaf-level stress tolerance and the two first axes of the phylogenetically corrected PCA built of the functional traits.Panel (a) shows the relationship between heat tolerance and PC1, (b) between drought tolerance and PC1 and (c) between drought tolerance and PC2.Points are mean values per species.Lines are predictions from linear regressions with ±95% confidence intervals.Only significant (p < 0.05, Table

Figure S1 :
Figure S1: The response of photosystem II function with increasing temperature for each species.

Figure S2 :
Figure S2: PCAs of the stress tolerance metrics and functional traits.

Figure S3 :
Figure S3: Comparison of minimum temperature of the coldest quartier and other early spring temperatures.

Figure S4 :
Figure S4: Relationship between mean last spring frost date and the other minimum temperature variables.

Figure S5 :
Figure S5: All pairwise correlations between the different stress tolerance metrics.

Figure S6 :
Figure S6: The response of photosystem II function with increasing temperature for two species with differently shaped temperature responses.

Figure S7 :
Figure S7: All pairwise correlations between stress tolerances and the PC axes built on functional traits.

Figure S8 :
Figure S8: Correlation matrix showing the relationship between directly measured stress tolerance and stress tolerances calculated from species climatic niche at their 10 or 1 percent quantiles.

Figure S9 :
Figure S9: Relationships between the most closely related stress tolerance metrics measured experimentally and from species climatic niches.

Figure S10 :
Figure S10: Relationships between the most closely related stress tolerance measures separated for conifers and broadleaves.

Figure S11 :
Figure S11: Normalized variability of LSF for each of the grid points in a 2.5° grid over Europe.

Figure S12 :
Figure S12:The Fv/Fm values before and after whole leaves were exposed to a heat treatment.

Figure S13 :
Figure S13: The Fv/Fm values of seedlings before and after whole seedlings were exposed to a heat treatment.File S1: Additional details on methods.File S2: Additional experiments on heat tolerance.

Table S8 :
Comparison of the experimentally measured tolerances and functional traits with descriptions of species' life history based on literature.