Recovery of Abies alba and Picea abies saplings to browsing and frost damage depends on seed source

Abstract The density of wild ungulates has increased in the last century, and browsing has become a major driver of forest succession in the northern hemisphere. In addition, tree species are expected to respond differently to future climate conditions, especially an increased frequency of late frost events. The aim of this study was to analyze the influence of intraspecific genetic variation on the recovery of two tree species to frost and browsing. An experiment with saplings from 90 Abies alba and 72 Picea abies seed sources was conducted. Five‐year‐old saplings were clipped at three intensities before budburst in spring. Growth (height, diameter, leader shoot length, and biomass) and quality (e.g. stem form, multistemming, reaction type) were assessed before and 1–2 years after clipping or 3–4 years after natural frost events, and provenance differences were related to environmental differences at the seed source. For Abies, frost and clipping resulted in reduced height growth in the first year after the stress and reduced height for two (clipping) to four (frost) vegetation periods. Sapling biomass, diameter increment, and quality decreased after heavy clipping. For Picea, which grew twice as fast as Abies, such effects were only found after frost damage. Population differences were significant for both species for all investigated growth traits and for Picea also for some quality variables. The “reaction type” after browsing (e.g. new shoot, existing twig bending upward) seems to be species specific and independent of seed source. In contrast, the time lag between clipping and formation of a clear new leader shoot increased for Abieswith lower temperatures at the seed source. Lowland populations with warmer climates grew faster, and for Picea also qualitatively better, and recovered faster from leader shoot loss (Abies) or reacted at the uppermost meristem (Picea). Thus, the investigated stressors increased the existing differences among populations.


| INTRODUC TI ON
There is growing evidence of increasing temperatures and decreasing summer precipitation in Central Europe (IPCC, 2013).
Climate change is likely to affect trees on many sites (Westerling, Hidalgo, Cayan, & Swetnam, 2006). Although the expected increase in the frequency of drought periods or late frost events is expected to affect individual tree species differently (Richter et al., 2012). However, the existence of intraspecies differences is also well known from provenance trials. Apart from differences in height growth, second flushing and phenology (Frank et al., 2017;Szeligowski, Bolibok, Buraczyk, & Drozdowski, 2011), considerable differences have been observed between populations in frost resistance (Hansen & Larsen, 2004) and drought tolerance (Csillery, Ovaskainen, Sperisen, Widmer, & Gugerli, 2018). For this reason, populations that do well under future climate conditions would be beneficial for forestry purposes.
However, tests should be completed to determine if ungulate browsing does not interfere with growth to such a degree that it counters the growth benefits of climatically better-adapted populations. In the last century, the density of chamois, roe, and red deer has increased (Apollonio, Andersen, & Putman, 2010). Therefore, ungulate browsing has become a major driver of forest succession in the northern hemisphere and can challenge the establishment of future tree generations (Tanentzap et al., 2009;Tremblay, Huot, & Potvin, 2007). In the face of increasing forest regeneration problems due to these high ungulate densities (Ramireza, Jansenb, & Lourens, 2018), it would be valuable to have better knowledge on the following: (a) the selective browsing on certain populations and (b) the recovery potential of different tree populations following browsing.
It is well known that browsing is inter-and intraspecifically a very selective process. For example, vigorously growing saplings are preferentially selected by ungulates (Iason, Duncan, Hartley, & Staines, 1996;Kupferschmid, 2018). Population differences in bud break and growth cessation have been shown to cause large differences in moose browsing in Finland (Viherä-Aarnio & Heikkilä, 2006). Clear differences in browsing frequency have also been found between populations of Eucalyptus morrisbyi trees (Mann et al., 2012). Apart from traits that help a plant avoid browsing selection, those that enable a plant to recover can be seen as a strategy to mitigate browsing impacts (O'Reilly-Wapstra et al., 2014).
Very few studies have been conducted to investigate the population differences in recovery following browsing. The growth trait differences in five populations of Pinus sylvestris were found to be maintained irrespective of browsing (O'Reilly-Wapstra et al., 2014), and the same was found for saplings of 77 Fagus sylvatica seed sources (Frank, Heiri, & Kupferschmid, 2019). However, the compensation capacity of tree species in response to leader shoot browsing depends on many factors, among them the architecture of the tree saplings (cf. review by Kupferschmid, 2017).
Therefore, predictions of no genetic differences in recovery from browsing for all tree species based on findings from P. sylvestris and F. sylvatica alone cannot be considered reliable.
In this study, we were interested in the population-specific reaction to browsing and the variation in recovery from leader loss of two commercially very important tree species of Central Europe, that is, Picea abies (L.) Karst. and Abies alba Mill. We knew from the results of common garden experiments that Picea has considerable growth differences between populations and that Abies has at least some differences (Csillery et al., 2018;Frank et al., 2017). Further, saplings of A. alba are browsed much more frequently and heavily by ungulates than saplings of P. abies (Kupferschmid, 2018;Vacek et al., 2014), which could have led to different genetic selection and thus to differences in the genetic variation in the recovery following browsing. Additionally, recent browsing experiments have already led to recommendations to foresters that A. alba populations producing many buds should be planted at sites with high ungulate density based on the assumption that bud formation of saplings depends on genetically fixed traits (Kolly & Kupferschmid, 2014). Due to the results obtained by Pinus and Fagus, it remains unclear to what extent, if at all, the recovery of Abies and Picea saplings after different intensities of browsing depends on seed origin. The aim of this study was thus to analyze the influence of intraspecific genetic variation in the recovery of two tree species to stress caused by simulated winter browsing and frost damage. The specific research questions were as follows: • How do A. alba and P. abies saplings react to simulated browsing and are their reactions dependent on population differences?
• Do different populations recover differently following leadershoot loss and thus show genetic variation in sapling recovery traits?
• Are there correlations between climatic and edaphic conditions at the seed source and population differences in the recovery following simulated winter browsing or frost damage?
• Are the genetic differences in the growth of A. alba and P. abies saplings maintained in the presence of light and heavy browsing?

| Plant material and experimental setup
In 2009, seeds were sampled from 72 P. abies (referred to as Picea) and 90 A. alba (referred to as Abies) seed sources, covering the entire range of climatic conditions suitable for each species in Switzerland (e.g., elevations from 400 to 2,000 m a.s.l., Frank et al., 2017). For each seed source, three parent trees were selected from the same stand but at least 100 m apart to minimize relatedness (Arnold et al., 2010). In April (A. alba) and May (Picea abies) 2010, seeds from each mother tree (referred to as a "family") were sown directly into nursery beds at the Swiss Federal Institute for Forest, Snow, and Landscape Research WSL in Birmensdorf, Switzerland.
Picea seeds were originally sampled from 92 seed sources, but 20 seed lots consisted of mixed seeds from 10 trees per seed source (Frank et al., 2017) and were omitted from our study. In addition, the seeds from twenty mother trees did not germinate properly, that is, for both species, eight seed sources were represented by seedlings of two families and two seed sources by seedlings of one family only. Throughout the paper, the term "population" refers to individuals whose seeds were collected at the same place of origin. The term "seed source" refers to the location of a population origin.
F I G U R E 1 Growth traits of Abies alba saplings before (2014), one (2015) and two (2016) vegetation periods after simulated browsing; "no" = not clipped and not damaged, "light" = only apical bud removal on the terminal shoot but no further damage, "heavy" = leader shoot and all twigs clipped but no further damage, "damaged" = frost damage before clipping. Lower case letters indicate significant differences at p ≤ 0.05 between the treatments in the Tukey post hoc tests (for cases where the variable "treatment" was significant at p < 0.01 in the linear mixed-effects models; see Table 1). Median (bold line), first, and third quartile (bottom and top of box), quartile ±1.5 * interquartile range (whiskers) and individual points more extreme in value (circles) were drawn using the "boxplot" function in default R code. The width of the boxes represents the number of trees within the various categories An extensive common garden (half-sib progeny) experiment was carried out at the study site Brunnersberg, a former pasture on a south-facing slope (20%-24% incline) in the Jura Mountains in Switzerland (47°19′35″N, 7°36′42″E, 1,090 m a.s.l.). The site is characterized by a mean annual temperature of 6°C, a mean annual precipitation sum of 1,400 mm (Frank et al., 2017), and a shallow rendzic soil.
In spring 2012, seedlings were transplanted to the study site as bare-rooted seedlings. The experimental design consisted of 16 plots per species (32 plots total), each plot with six rows of seedlings and a spacing of 30 cm × 40 cm between the seedlings in each plot.
Each plot contained one seedling per family, that is, mostly three individuals per population, randomly distributed within the plot. For a detailed description of seed collection, common garden procedures, and the random block design, we refer to Frank et al. (2017). Height, basal diameter, bud phenology, and growth duration were measured in spring and autumn 2013, and values were reported by Frank et al. (2017).

| Environmental variables at seed sources
Environmental variables considered at the seed sources included the following: mean annual temperature (MAT); mean spring temperature (March -May, MTsp); continentality (interannual temperature variance, that is, maximum of warmest month minus minimum of coldest month); average maximum diurnal amplitude of temperature during spring (March -May, DTAsp); sum of growing degree days (based on a threshold of 5°C, DDEG); average numbers of days during the vegetation season (March -November) with frost (SFROv); mean annual precipitation sum (PREC); absolute maximum drought (PREC < 0.01 mm) period length in summer (June -August, DRYPsu); and annual aridity index (DMI = PREC/MAT*10 (Martonne, 1926)). All variables were calculated for the period 1931-1960 for each seed source (Frank et al., 2017). Physical and chemical soil properties-including the available water capacity of 1 m soil depth (AWC)-were derived from local soil pits that were located within a few meters of one of the three mother trees at each seed source (details see appendix in Frank et al., 2017). Note. Model details include the transformation (TF) applied to response variables and covariates, the covariate included (D12 = basal diameter in 2012, H12 = height in 2012), the number of analyzed saplings (N tot), and the mean and standard deviation (SD) of the trait (not transformed). Trait "DI" is the diameter increment measured as diameter in 2016 minus diameter in 2014. For the random effects, the variances are given, and for the fixed effects, the estimated coefficients are given. Population has been cut to Pop and the treatments (T) to "no" for no clipping and not damaged, "L" for light clipping, "H" for heavy clipping and "D" for damaged. p values of the likelihood ratio tests for each variable in the full models (DF = 10) and of the Tukey post hoc tests of the treatment are printed in bold and italics for p ≤ 0.01 and in italics for p ≤ 0.05.

| Frost damage
In all plots, several saplings had been damaged by late frost events in spring 2013 and 2014. The distribution of the frost events was not even between the plots. However, apart from the exceptions mentioned above, three seedlings of each seed source were planted in every plot, and the position within the plot was randomized.

| Trait assessment
The growth and quality of saplings were assessed before and one to two vegetation seasons after clipping (Supporting Information Table   S1 and Table S2). Sapling height was measured as the vertical distance from the ground surface to the tip of the leader shoot (height) or to the highest point of the tree regardless of whether this was a leader shoot or a branch (tree height). The annual height increment of the leader shoot was measured to an accuracy of 0.5 cm. Measurements of stem diameter were taken 2 cm above the soil surface using electronic calipers (Type M-150, MBFZ toolcraft GmbH, Georgensgmünd, Germany). In February 2017, all Abies were cut 2 cm above the soil surface and their fresh weight was determined immediately with a hanging scale (Kern HDBH 5K5N) with a resolution of 5 g.
In order to estimate dry weight, 50 saplings of each species were harvested from two control plots at the end of the ex- (deviation from vertical line ≤22.5°), "bent" (deviation 22.5-45°), or "severely bent" (deviation>45°). Crown form was classified using five levels ranging from optimal (1) to low quality (5; a detailed field guide will be made available on EnviDat). In addition, the vitality of living saplings in 2016 was assigned to one of five classes ranging from vital (0) to the presence of several completely withered branches (4).
In autumn 2016, we recorded how the sapling had reacted to leader loss due to frost damage or clipping by evaluating the "location of reaction," the "reaction type" and the "time lag" of the reaction.
The "location of reaction" had three levels: reaction out of "uppermost shoot whorl," reaction out of "lower shoot whorl," and "no reaction." There were six "reaction types": (a) production of a "new distal shoot" out of a bud on the stem or on the remaining leader shoot pieces, (b) production of a "new basal shoot" out of a bud on a whorl, (c) "flagging" of an existing internodal side shoot, (d) "flagging" of an existing whorl shoot, (e) use of an already vertically growing or bending upward of an "epicormic shoot" that is, a shoot that developed before the stress but was younger than the regular whorl shoots (preventitious shoots after Gruber, 1994), and (f) "no reaction." The time lag of the reaction after clipping was evaluated as: "0" = clear new leader shoot (longer than 0.5 cm) formed in the first vegetation season after leader loss, "1" = new leader shoot formed in 2016, and "2" = no new leader shoot formed until the end of the second growing season. Likewise, the time lag of the reaction after frost damage (0-4 years) was noted.
We recorded whether the saplings made no second flush at the 2016 leader shoot (0) Christensen, 2015). For the binary trait "frost damage occurrence,"

| Data analysis
we applied a generalized linear mixed-effects model using the R function glmer (package "lme4," binomial model, link = "logit," optimizer = "bobyqa"). An intercept, the treatment (4 levels: no, light and heavy clipping, frost damage) and a covariate (see below) were included as fixed effects and plot, population, family, and the interaction between population and treatment were included as random effects in the mixed-effects models. The interaction between population and treatment was not significant and including it resulted in model convergence failure. This interaction was thus only retained in the model for the continuous traits of Abies. The covariate (omitted for the binary trait "frost damage") was either height or diameter in 2012, and it was included to account for differences already apparent in the nursery and differences caused by different planting depths.
To test for the significance of all factors in our mixed-effects model, likelihood ratio tests were used that compared the full model to the same model without the terms that should be tested (R function ANOVA). If the treatment was significant at p < 0.01, a Tukey post hoc test was used to distinguish between the effects of the four treatment levels, that is, the three clipping levels and (frost) "damaged" (R function glht, package "multcomp"; (Hothorn, Bretz, & Westfall, 2008)).
In cases where there was a significant population effect (p < 0.01) in the mixed-effects model, we calculated Pearson correlations using the R function rcorr (package "Hmisc") between population random effects obtained from the analyses of variance (R function ranef, package "lme4") and site-specific environmental variables at the seed source. If a correlation was significant at p < 0.01, we used robust line fitting (R function line) to analyze the linear relationships between the population random effects of the respective trait with this site-specific environmental variable.
To investigate the effect of "reaction type" on the height of the saplings in 2016, similar linear mixed-effects models were fitted but "reaction type" was added as a fixed effect instead of treatment. These models were fitted separately for lightly and heavily clipped saplings (using the R function lmer) and a Tukey post hoc test was used to distinguish between the effects of the "reaction types".

| Abies alba
Frost damage and clipping both resulted in reduced height growth of Abies in the first year after the stress but not in the following years ( Figure 1). No significant difference was found between light and heavy clipping. Height was reduced for at least two (clipping) and up to four (frost damage) vegetation seasons in comparison with F I G U R E 2 Morphological traits of Abies (a) and Picea (b) saplings two vegetation periods after simulated browsing; "no" = not clipped, "light" = only bud removal, and "heavy" = leader shoot clipped and in the case of Abies alba also all twigs clipped, "damaged" = frost damage before clipping. Lower case letters indicate significant differences at p ≤ 0.05 between the treatments in the Tukey post hoc tests (for cases where p ≤ 0.001 in the mixed-effect models; see Tables 2 and 5) unclipped and undamaged trees. Height at the tree top was greater for lightly than for heavily clipped trees, as branches of lightly clipped Abies were often higher than the new leader shoot. Basal diameter was affected neither by frost damage nor by clipping (Table 1)  "Reaction type" and "location of reaction" differed between the treatments (Table 2) (Table 1). Of all Abies saplings, 28.4% showed a reaction time lag of one year, that is, they mostly formed a new visible bud without elongation growth. About 4.9% of the saplings (108 of the 2,187 saplings) still had no leader shoot at the end of the experiment.
Population differences were significant for all growth traits except the diameter increment (Table 1), but hardly ever for quality variables (Table 2). Diameter, height, height increment, and aboveground biomass all decreased with decreasing temperature (MAT and MTsp; Figure 4a), DDEG and continentality at the seed source (Table 3). Therefore, these traits also decreased with increasing elevation, geographical longitude, and hill slope. Fewer frosts (SFROv) and a drier climate at the seed source (PREC, DryPsu, DMI) corresponded to higher values of most growth traits. Soil properties had less influence than the climate at the seed source, but less sand, more clay and a smaller C/N ratio were correlated with higher values of many growth traits (Table 3). The number of buds on the leader shoots formed in 2014 increased with increasing temperature at the seed source (Table 3) and was positively correlated with diameter and height in 2014 (Pearson correlation coefficient (corr) of 0.6).
Neither the "reaction type" nor the "location of reaction" differed among the provenances, but the time until a new leader was formed varied significantly among the Abies populations ( Table 2) Figure S1), independent of seed source (Table 2).
However, seed source nevertheless seemed to play a role in the reaction to frost, as the "reaction type" after frost varied among the families (Table 2)

| Picea abies
Picea grew more than twice as fast as Abies saplings in both height and stem diameter (Figure 1 vs. Figure 5). Frost damage but not clipping resulted in a significant reduction of height growth, height, biomass, and basal diameter of Picea for at least three to four vegetation seasons ( Figure 5). Nevertheless, lightly clipped Picea had larger height increments in the year 2016 compared to heavily clipped saplings (Table 4).
Most quality traits were significantly reduced for the clipped and "frost damaged" Picea. Multistemming very rarely occurred for control saplings, was frequent for lightly clipped saplings, and was common for heavily clipped and frost damaged saplings (Figure 2b).
Crown form had a reduced quality in clipped and frost damaged saplings, irrespective of the cause of leader loss. Stem form and sapling vitality were not affected by clipping or frost damage (Table 5).
All Picea had a new leader shoot at the end of the experiment ( Figure 3b). Some heavily clipped Picea (5.2%) reacted with an "unclear" new leader shoot in the first year after clipping, owing to side shoots that were not fully bent upward. "Reaction type" but not "location of reaction" differed between light and heavily clipping treatments (Table 5)  at the seed source, more Picea had an early prolepsis with a second flush without bud dormancy and then a late prolepsis with a third flush after bud dormancy. However, the 632 saplings with clear elongation growth of their proleptic leader shoot (longer than 2 cm) did not differ among treatments or populations regarding the total length of the proleptic shoot (Table 4).
The "reaction type," time lag and multistemming after clipping were not influenced by population differences (Table 5). However, the buds or branches of the new shoots originated lower down the stem with a higher DMI, decreasing continentality (Figure 4b), smaller DTAsp, lower soil pH, less clay but more sand in the soil, and steeper slope at the seed source (Table 6).  Table 5). The model for "frost damage occurrence" did not reveal any F I G U R E 3 Relationship between height in 2016 and reaction type for Abies (a) and Picea (b) saplings that were not clipped (control), lightly clipped or heavily clipped, excluding all frost damaged trees. The number of observations (N) per reaction type are shown in gray. Lower case letters indicate significant differences at p ≤ 0.05 between the reaction types within each treatment in the Tukey post hoc tests significant differences between the provenances or families of Picea (p = 0.1519).

| D ISCUSS I ON
Knowledge is needed about genetic influences on multiple species to form recommendations in the face of climate change and under the current high ungulate pressure. We analyzed intra-specific differences of two of the most important tree species in Switzerland (Cioldi et al., 2010), which differ in their selection by wild ungulatesthat is, A. alba is selected much more often than P. abies (Vacek et al., 2014).

| How do A. alba and P. abies saplings react to simulated browsing and are their reactions dependent on population differences?
There were intrinsic differences in the reaction to simulated browsing between the two species. Almost no Abies reacted with F I G U R E 4 Linear relationships of population effects to environmental conditions at seed source for 90 Abies (a) and 72 Picea (b) populations from Switzerland. Note that the population effects are not equally scaled. Regression lines are displayed only for significant (p < 0.001) relationships in the linear mixed-effect models (Tables 1,2 and 4,5) and in the Pearson correlations (Tables 3 and 6). Lines are drawn using robust line fitting (R function line). In the case of Abies alba and time lag, regression lines are shown separately for lightly (dashed black) and heavily (solid gray) clipped saplings the bending upward of a previously existing branch, while for Picea flagging was very common after light and heavy clipping.
Generally, flagging seems to be an efficient but rather rare reaction type for Abies saplings , in particular under natural browsing (Kupferschmid, Wasem, & Bugmann, 2015). The different reactions of the species to clipping could be caused by the plagiotropic growth of Abies but not Picea branches; that is, Abies follows the architectural model MASSART and Picea follows RAUH (Hallé & Oldeman, 1970). Plagiotropic growth of branches is said to be genetically determined (Bartels, 1993). Based on the results of our study, this genetic effect seems to operate on a species level rather than a within-species level, as the population effect for "reaction type" was not significant for either species.
The production of a "new basal shoot" out of a bud on the uppermost whorl was the most frequent reaction of Abies to heavy simulated browsing and frost damage. Epicormic shoots were also more frequent for Abies than for Picea. We did not analyze if these basal shoots were formed from small axillary buds at the base of the whorl branches that had remained dormant because of apical dominance or if they were adventitiously formed buds (Meier, Saunders, & Michler, 2012). In any case, these buds were not visible by eye before browsing. Tree vigor may play a major role in the formation of epicormic shoots, with less vigorous trees forming more such shoots, especially after pruning (Meier et al., 2012). Abies has a greater potential for forming epicormic shoots because it has more and longer-living dormant (inactive according to Bonser & Aarssen, 1996) meristems compared with Picea (Gruber, 1994). Some evidence was found that differences in epicormic branch production are based on population-level heritability (Meier et al., 2012). We found popu- faster growth of Picea, which resulted in Picea saplings that were twice as tall as Abies saplings, it is not possible to determine whether the recovery of these two species differed. The very shade tolerant, deep rooting Abies were clearly more stressed than Picea at our fully sun-exposed site on shallow soil, and this may have been the reason for the smaller growth and relatively poor recovery of Abies.
One of the weaknesses of our experiment was that we had no replication. Our results would be more informative if the experiment had been carried out at two or more locations or even with a fully reciprocal experimental design. However, no interactions between treatment and site were found for morphological traits regarding the recovery of Fagus sylvatica saplings when results from our study sites were compared with those at the lowland study site Birmensdorf (Frank et al., 2019). In addition, a reciprocal common garden experiment can be performed with only a few populations and not with seeds from 90 Abies and 72 Picea seed sources. Another limitation of our experiment is that it was carried out with only a single generation and without knowing the specific genotypes of the populations. We thus interpret our results with caution and recommend selecting specific populations of particular interest for conducting further in-depth analysis under more natural-that is, at least partly shaded-conditions.

| Do different populations recover differently following leader shoot loss?
In our study, the reaction type had an influence on the capacity to recover following browsing ( Figure 3). However, reaction type did not depend on population. This suggests no direct involvement of the different seed sources in the recovery via the reaction type. For both species, however, population-related differences regarding recovery were found, that is, differences in the time lag for Abies and in the location of reaction for Picea.
One-third of the Abies saplings only formed a new bud and not a real new shoot in the first years after clipping. Thus, the reaction was delayed for one or two years for Abies and the saplings were not able to recover the height loss. Such time lags in reaction to browsing for Abies have already been detected in many studies (Kupferschmid, Zimmermann, & Bugmann, 2013). There was a significant population effect on the time lag, and it occurred irrespective of treatment intensity. This suggests that the time lag for Abies has a heritable basis and is not just due to light or heavy browsing.
To our knowledge, this is the first time population-related differences in the delay in the reaction to clipping, and thus in the recovery, have been shown. Reactions to clipping were immediate for Picea and thus independent of seed source. However, the "location of reaction" differed among Picea populations at our site (see below).

| Are there correlations between climatic and edaphic conditions at the seed source and the population differences in the recovery following browsing?
In our study, Abies populations from lowland locations with a warmer climate at the seed source reacted faster after clipping than highland populations from colder climates (Figure 4a). However, lowland populations were also taller, thicker and had more aboveground biomass. Evergreen conifers have previously been found to retain reserves, particularly in the youngest age class of needles (Millard, 1995). It is therefore likely that the larger lightly clipped Abies had more needles available from which reserves for a reaction could be mobilized. Thus, our results support earlier findings that the more vigorously an Abies is growing, the faster it reacts after simulated browsing . The vigorous growth of Abies additionally seems to be dependent on seed source (Table 1; Hansen & Larsen, 2004;Szeligowski et al., 2011). Albeit, the population differences occurred for lightly and heavily clipped Abies and tended to be even larger for heavily clipped saplings that had all their youngest needles cut (Figure 4a). Thus, it is likely that, among others, a population-dependent mechanism controls the time needed for a reaction and hence the recovery following leader-shoot loss in Abies.
For Picea, the "location of reaction" differed among populations, in that trees reacted more efficiently by using the uppermost possible buds or branches when they came from regions with more climate variation over the year (Figure 4b). As "location of reaction" correlated negatively with diameter, height, and aboveground biomass (Pearson correlation around −0.2), it seems that more vigorously growing Picea reacted more efficiently. For Picea, this vigorous growth is at least partly inherited (Table 4; Frank et al., 2017), as population differences in height and diameter have not been found on all sites (Burger, 1941).

| Are the genetic differences in the growth of A. alba and P. abies saplings maintained in the presence of light and heavy browsing?
First, we were able to confirm that population differences occur for both species in the absence of browsing (e.g. results for the 2014 traits). Second, we tested the hypothesis ungulate browsing does not interfere with growth to such a degree that it counters the growth advantages of the most vigorous populations at our site.
We found no evidence that this was the case. In contrast, browsing increased the population differences in height, diameter, and biomass. Note. Details as in Table 1.
For Abies, the main reason for this finding is that many slowgrowing Abies saplings from colder high-elevation seed sources needed one or more years to form a new real leader shoot at our site. This corresponds well with the fact that the often-observed regeneration failure of Abies is more accentuated in mountain than lowland forest in Switzerland (e.g. the tolerable browsing limits were derived for mountain forests, cf. Eiberle & Nigg, 1987).
For Picea as well, the population differences have increased for all growth traits since the onset of the experiment. The recovery of Picea saplings after this single clipping event was good overall, and we would expect major population differences after repeated browsing, owing to the larger influence on the traits (see review of simulated clipping experiments in Kupferschmid, 2017), and after frost damage.

| Frost damage versus browsing
Late spring frost can cause equally severe (Abies) or even worse (Picea) damage to tree saplings than the browsing simulated in our experiment. Populations with poor winter-frost resistance can have high mortality rates at sites where late frosts occur (Hansen & Larsen, 2004). Our results suggest that populations of Abies from warmer and more humid seed sources may be more sensitive to frost than populations with colder and drier conditions at the seed source. Larsen (1986) found particularly large variations in the fast-growing Calabrian populations of A. alba, with increasing frost resistance occurring with increasing elevation. It also seems that high-elevation populations of Picea have greater frost resistance than lowland populations because they have fewer proleptic shoots (Gruber, 1994), but we had too few frost damaged Picea to analyze this aspect. Thus, lowland populations are probably overall more prone to leader damage by frost.

| CON CLUS IONS
Based on common garden experiments with A. alba, some authors have concluded that there is no need to select for specific populations, as Abies is a very adaptable species (Frank et al., 2017;Vitasse, Delzon, Bresson, Michalet, & Kremer, 2009). We found somewhat smaller population differences for Abies than for Picea saplings, but 17 out of 28 variables still showed important population effects on the Abies saplings in our study (Tables 1 and 2). In addition, for three-to six-year-old saplings, the temporal trend of heritability estimates for total height were found to increase with age in one study (Mihai, Mihaigmihai, & Duta, 2014). However, differences in height after six growing seasons were found to be much larger than after 46 growing seasons in another study (Kerr, Stokes, Peace, & Jinks, 2015). Thus, the population differences in growth and quality We found that the existing differences among populations increased because of differences in the capability of saplings to recover growth after a frost event or simulated browsing. Lowland populations from warmer climates grew faster, and for Picea also qualitatively better, and recovered faster following leader shoot loss (Abies) or reacted with new growth at the uppermost meristem (Picea). Thus, even single browsing events can hamper the growth of trees, at least Abies saplings, but populations with fast growth can be expected to react rapidly and efficiently to leader shoot damage.
Browsing recovery should be incorporated into forest manager decisions regarding the strategy for regeneration, especially in heavily browsed areas. The interactive effects of site conditions, seed source, and population differences in recovery to stress caused by browsing and frost should be considered carefully. Taking these effects into account could make an important F I G U R E 5 Growth traits of Picea abies saplings before (2014), one (2015) and two (2016) vegetation periods after simulated browsing; "no" = not clipped, "light" = only apical bud removal on the terminal shoot, "heavy" = leader shoot clipped, "damaged" = frost damage before clipping. Letters refer to significant differences at p ≤ 0.05 between the treatments in the Tukey post hoc tests. For plotting details, see Figure 1 and thus cost-effective contribution to ensuring that our forests steadily provide their ecosystem goods and services, such as protection from natural hazards, biodiversity preservation, and timber production.

ACK N OWLED G M ENTS
We thank the ADAPT project team at the Swiss Federal Research Institute WSL for granting us access to their field site for this study.
We are grateful to several field assistants for measuring thousands of saplings. We thank Jens Nitzsche and his teams from WSL for tending the site. M. Dawes is gratefully acknowledged for providing linguistic suggestions. We are grateful to two reviewers for their comments on an earlier version. This study was funded by the

CO N FLI C T O F I NTE R E S T
None declared.

AUTH O R S' CO NTR I B UTI O N S
AK and CH conceived the ideas, acquired funding, and designed methodology. CH was responsible for the acquisition of data before clipping, while AK collected all data during and after clipping. AK analyzed the data, and both authors interpreted the results. AK led the writing of the manuscript. Both authors contributed critically to the drafts and gave final approval for publication.

DATA ACCE SS I B I LIT Y
Data from the clipping experiment will be archived at EnviDat with a DOI.