Risk of Genetic Maladaptation Due to Climate Change in Three Major European Tree Species

Received: 27 February 2017 | Accepted: 13 June 2017 DOI: 10.1111/gcb.13802

PRIMARY RESEARCH ARTICLE

Risk of genetic maladaptation due to climate change in three major European tree species

1Swiss Federal Institute for Forest, Snow and Landscape Research WSL, Birmensdorf, Abstract Switzerland Tree populations usually show adaptations to their local environments as a result of 2Department of Forest Ecosystems and natural selection. As climates change, populations can become locally maladapted Society, Oregon State University, Corvallis, OR, USA and decline in fitness. Evaluating the expected degree of genetic maladaptation due 3Pacific Northwest Research Station, USDA to climate change will allow forest managers to assess forest vulnerability, and Forest Service, Corvallis, OR, USA develop strategies to preserve forest health and productivity. We studied potential Correspondence genetic maladaptation to future climates in three major European tree species, Nor- Aline Frank, Swiss Federal Institute for Forest, Snow and Landscape Research WSL, way spruce (Picea abies), silver fir (Abies alba), and European beech (Fagus sylvatica). Birmensdorf, Switzerland. A common garden experiment was conducted to evaluate the quantitative genetic Email: [email protected] variation in growth and phenology of seedlings from 77 to 92 native populations of each species from across Switzerland. We used multivariate genecological models to associate population variation with past seed source climates, and to estimate relative risk of maladaptation to current and future climates based on key phenotypic traits and three regional climate projections within the A1B scenario. Current risks from climate change were similar to average risks from current seed transfer practices. For all three climate models, future risks increased in spruce and beech until the end of the century, but remained low in fir. Largest average risks associated with climate projections for the period 2061–2090 were found for spruce seedling height (0.64), and for beech bud break and leaf senescence (0.52 and 0.46). Future risks for spruce were high across Switzerland. However, areas of high risk were also found in drought-prone regions for beech and in the southern Alps for fir. Genetic maladaptation to future climates is likely to become a problem for spruce and beech by the end of this century, but probably not for fir. Consequently, forest management strategies should be adjusted in the study area for spruce and beech to maintain productive and healthy forests in the future.

KEYWORDS Abies alba, climate change, Fagus sylvatica, genecology, local adaptation, Picea abies, quantitative traits, relative risk of maladaptation, seedling common garden experiment, water availability

1 | INTRODUCTION of local maladaptation to future climates for populations and species. Relative risk quantifies the difference between populations adapted Tree species of temperate and boreal regions often exhibit multiple to two different climates, e.g., past and future climates, taking into genetic adaptations to their local climates (Alberto et al., 2013; Bus- account the amount of within-population genetic variation (Camp- sotti, Pollastrini, Holland, & Bruggemann,€ 2015). For example, the bell, 1986; St. Clair & Howe, 2007). The quantitative genetic statis- timing of bud break and bud set typically varies along latitudinal and tics and climate associations needed to calculate relative risk of elevational gradients, and drought tolerance appears to be higher in climate change can be obtained from common garden experiments populations from dry environments. Such climatic adaptations are (e.g., St. Clair, Mandel, & Vance-Borland, 2005). In addition, high- considered to result from diversifying natural selection (Savolainen, resolution climate projections are needed, particularly changes in Pyh€ajarvi,€ & Knurr,€ 2007). As local climates change, however, tree temperature and precipitation. species may become maladapted if evolutionary adaptation does not In this study, we focused on Norway spruce (referred to as keep pace with ongoing environmental changes (e.g., St. Clair & “spruce”), silver fir (“fir”), and European beech (“beech”, Fagus sylvat- Howe, 2007). The resulting genetic maladaptation could lead to ica L.), three major European tree species whose ranges partly over- reduced fitness or even local extinction of current tree populations. lap in Central Europe (EUFORGEN, 2009). The climatic conditions in This has the potential to affect forest composition, structure, and Central Europe are expected to change markedly by the end of the stability, with potential negative consequences for the provision of century (2051–2080) compared to the second half of the 20th cen- forest goods and services (Lindner et al., 2010). tury (1951–2000), with mean summer temperatures increasing Different levels of climate-induced maladaptation are expected between 1.3 and 2.7°C, and summer precipitation decreasing by up to occur in different tree species. Adaptive specialists, such as lodge- to 25% (Lindner et al., 2014). The impact of climate change will likely pole pine (Pinus contorta), Douglas-fir (Pseudotsuga menziesii), and vary among regions and locations, and is considered to be especially Norway spruce (Picea abies [L.] Karst.), show high levels of climate- pronounced in mountainous areas such as Switzerland (Pepin et al., related population differentiation. They are likely at higher risk of 2015). Here, under the A1B scenario, mean summer temperatures maladaptation than are adaptive generalists such as white pine (Pinus are projected to increase by more than 4°C by the end of the 21st monticola), western redcedar (Thuja plicata), and silver fir (Abies century compared to the period 1980–2009 (CH2011, 2011). The alba Mill.) (Frank, Sperisen, et al., 2017; Rehfeldt, 1994; St. Clair & expected changes in temperature and precipitation may affect Howe, 2007). At the population level, the degree of maladaptation growth, fitness, and the distribution of all three tree species (Gessler might vary across the landscape, depending on the amount of et al., 2007; Hanewinkel, Cullmann, Schelhaas, Nabuurs, & Zimmer- within-population genetic variation, environmental heterogeneity, mann, 2013; Lebourgeois, Rathgeber, & Ulrich, 2010; Meier, adaptational lag, and the local extent of climate change (St. Clair & Edwards, Kienast, Dobbertin, & Zimmermann, 2011; Nothdurft, Howe, 2007). However, estimates for climate-induced maladaptation 2013; Nothdurft, Wolf, Ringeler, Bohner,€ & Saborowski, 2012). between and within species are rare. Recent results from a seedling common garden study have shown Knowledge of trees’ maladaptation to future climates is valuable that spruce, fir, and beech in Switzerland are characterized by dis- for developing and refining forest management strategies and tools, tinct genecological patterns (Frank, Pluess, Howe, Sperisen, & Heiri, such as seed transfer guidelines, that could help to mitigate negative 2017; Frank, Sperisen, et al., 2017). Genetic clines for spruce are climate change impacts on forest ecosystems (Park et al., 2014). Tra- pronounced along temperature gradients, whereas for fir and beech, ditionally, seed transfer guidelines and seed zones have been used genetic clines are weaker, and are mostly found along gradients of to conserve or enhance forest productivity and timber quality (Lan- temperature and water availability. These genecological patterns sug- glet, 1971). Such guidelines should now be reconsidered to preserve gest vulnerability to climate change is larger for spruce than for fir forest health and productivity in potentially warmer and drier cli- and beech. However, quantitative estimates of maladaptation to climates. Forest managers may, for example, select seed sources that mate change are lacking for these and most other tree species (but match the future local climate of a particular forest site, and use see St. Clair & Howe, 2007). In particular, we have no information such “preadapted” plant material for reforestation or admixture about the differences in potential maladaptation between spruce, fir, within existing stands (a.k.a., assisted migration or assisted gene and beech, or the variation in risk across the landscape. Such infor- flow; Aitken & Whitlock, 2013; Williams & Dumroese, 2013). For mation, however, could form the basis for science-based recommen- that purpose, forest managers need to know which species and dations for forest management, particularly for establishing regions are most vulnerable to climate change, and which stands guidelines of climate change adjusted seed transfer. Therefore, we could serve as sources of reproductive material preadapted to future addressed the following main questions using a genecological climates. approach: (1) Are current populations of spruce, fir, and beech in We used “relative risk of genetic maladaptation” to assess the Switzerland genetically maladapted to climate change? (2) Which vulnerability of trees to climate change. This index, which was origi- species and regions are most vulnerable to future maladaptation, and nally developed by Campbell (1986) to evaluate the genetic risk of why? We then discuss potential forest management practices to populations due to seed transfer, can also be used to assess patterns maintain forest health and productivity in the future. 5360 | FRANK ET AL.

seedling (or by three seedlings for the pooled seedlots of spruce), 2 | MATERIALS AND METHODS and all seedlings were randomized within blocks. We used data on third-year seedling growth and phenology of 2.1 | Plant materials and common garden beech, and data on fourth- and fifth-year seedling growth and phe- procedures nology of spruce and fir (see Fig. S1, Appendix S1, for a description This study was based on plant materials and common garden pro- of the sowing, planting, and measurement schedule). These measure- cedures described by Frank, Pluess, et al. (2017) and Frank, Speri- ments were described by Frank, Pluess, et al. (2017) and Frank, sen, et al. (2017). Briefly, we sampled 92 populations of spruce, Sperisen, et al. (2017). We selected three comparable key pheno- 90 populations of fir, and 77 populations of beech from their nat- typic traits for each species that showed high among-population ural ranges in Switzerland (Figure 1). Only native, i.e., autochtho- variation and strong relationships to climate for at least one of the nous, populations were sampled, and we aimed at covering large three species. For spruce and fir, these traits were total height (H), environmental gradients. For most populations, seeds were col- the timing of bud break (BudBreak), and the timing of growth cessa- lected from three trees. Exceptions for spruce included 20 pooled tion (GrowthCess). For beech, we used H, BudBreak, and the timing seedlots and one population with ten sampled trees. Seed trees of leaf senescence (LeafSen). We analyzed individual traits and prin- were chosen to represent the overall characteristics of the stand cipal component (PC) scores, but only report the results for the indi- (i.e., with respect to aspect, slope, soil). Where possible, trees in a vidual traits. This is, because the PC scores did not produce stand were selected to be separated by at least 100 m to reduce genecological models with larger R2 compared to the individual traits, the likelihood of including close relatives. Decreasing relatedness and are more difficult to interpret (Table S1, Appendix S1). between individuals with increasing distance between trees was confirmed for beech in a parallel study using molecular markers 2.2 | Past and current climates (Pluess et al., 2016), but was not tested for spruce and fir. The progenies (open-pollinated families) were grown in the nurs- Past climate at seed sources was inferred using meteorological data ery at the Swiss Federal Institute for Forest, Snow and Landscape from 1931 to 1960, the period that most closely matched the estab- Research WSL in Birmensdorf, Switzerland. To obtain seedlings of lishment period of the seed trees used in this study. Although the comparable size, beech was cultivated in the nursery for 1 year, but ages of the seed trees are unknown, spruce, fir, and beech stands spruce and fir were grown for 2 years (Fig. S1, Appendix S1). Subse- typically do not reach sexual maturity until age 40 to 70 (Houston quently, the seedlings were transferred to a common garden field Durrant, De Rigo, & Caudullo, 2016; Modrzynski, 2007; Wolf, 2003). site located at Brunnersberg (47°19035″N, 7°36042″E) in the Jura However, based on their diameters, most trees were probably 80 to mountains at an elevation of 1090 m a.s.l. (Figure 1). There, the 150 years old in 2009 and 2011 when the seeds were collected. seedlings were planted in 16 blocks per species at 30 cm 9 40 cm Thus, we used the earliest climate data available (1931–1960) to spacing. Within each block, every family was represented by one represent the climate to which the tree populations were adapted to

FIGURE 1 Distribution of the 92 Norway spruce (Picea abies), 90 silver fir (Abies alba), and 77 European beech (Fagus sylvatica) populations sampled across Switzerland. A star indicates the common garden site. Colored regions represent the six main biogeographic regions of Switzerland (Gonseth et al., 2001). FRANK ET AL. | 5361 before current climate change. Past climate data were obtained from (2014): potential evapotranspiration (PET) according to Romanenko 21 climate stations that recorded air temperature (T) and dew point (1961), and site water balance (SWB) according to Grier and Running temperature (Td), and from 24 stations that recorded precipitation (1977). SWB is a function of P, PET, and plant available water capacity (P; Swiss Federal Office of Meteorology and Climatology Meteos- (AWC). For all seed sources, AWC had been specifically estimated wiss). These climate and precipitation stations provide the earliest from local soil profiles (Frank, Sperisen, et al., 2017; Teepe, Dilling, & reliable climate data that are available. The stations are mostly part Beese, 2003), whereas for the NFI sample plots, AWC was derived of the Swiss National Basic Climatological Network (Swiss NBCN), from a Swiss-wide AWC map (Remund & Augustin, 2015). We then and cover all 12 climate regions of Switzerland (Begert et al., 2007). used the daily estimates for P, T, and RH, and the monthly values for Current climate was inferred from meteorological data from PET and SWB to derive the climate variables shown in Table 1. The 1981 to 2000, obtained from 79 temperature, 71 dew point temper- interpolations were slightly better for current than for past climate ature, and 371 precipitation measurement stations from across due to the increasing number of climate stations. Standard deviations Switzerland (Remund, 2016). Past and current stations were almost from cross-validations were 1.6°C (T), 5.2 mm (P), 13.1% (RH), 1.0 mm equally distributed across Switzerland, with only some western parts (PET), and 60.4 mm (SWB) for 1931–1960, and 1.4°C (T), 3.8 mm (P), being less well covered in the past. 8.7% (RH), 0.7 mm (PET), and 47.2 mm (SWB) for 1981–2000 For the past and current time periods, daily mean values of T, Td, (Remund, 2016). Past climate changes were calculated as the differ- and P, and monthly values of P were available from the stations ences in mean climate between 1981–2000 and 1931–1960. described above. These data were spatially interpolated to the 259 seed source locations of spruce, fir, and beech, and to 13,581 sample 2.3 | Future climate plots located on a 1 km-grid that had been classified as “forest” in the Swiss National Forest Inventory (NFI; sample plots of first and second The future climates of the seed source locations and NFI sample plots survey). These interpolations were performed using the enhanced were projected using three selected regional climate models (RCMs), Shepard’s Gravity Interpolation method that accounts for the three- which represent the range of summer precipitation and temperatures dimensional distance between climate stations, and integrates local described in the CH2011 Swiss Climate Change Scenarios (CH2011, effects of lakes, cities, slope orientation, and elevation (Remund, Freh- 2011). Specifically, the three representative RCMs were selected ner, Walthert, K€agi, & Rihm, 2011; Remund, Rihm, & Huguenin-Landl, based on their projected climates for a northern and a southern loca- 2014; Zelenka et al., 1992). For each site, the nearest eight climate tion in Switzerland (north: Aarau, 47.38°N, 8.08°E, 394 m a.s.l.; stations were used for interpolation. Using the interpolated data, rela- south: Locarno, 46.17°N, 8.80°E, 223 m a.s.l.; Remund et al., 2014). tive humidity (RH) was derived from dew point temperature (DWD, This was done using the intermediate A1B emission scenario of the 1979). In addition, we used two variables calculated by Remund et al. fourth IPCC climate change report (AR4; Nakicenovic & Swart, 2000),

TABLE 1 Climate variables used to describe seed source climates of 92 Norway spruce (Picea abies), 90 silver fir (Abies alba), and 77 European beech (Fagus sylvatica) populations across Switzerland

Abbreviation Unit Description Models* Temperature†‡ MAT °C Mean annual temperature MTwarm °C Mean temperature of warmest month a MTcold °C Mean temperature of coldest month MTsp °C Mean spring temperature (March–May) b, c CD d Chilling days; number of days with average temperature ≤ 5°C CONT °C Continentality (inner-annual temperature variance) a, b, c Water availability†‡ PRCan mm Annual precipitation sum PRCsu mm Summer precipitation sum (June–August) a, b, c PRCwi mm Winter precipitation sum (December–February) a, b, c PRCPETveg§ mm Water balance (precipitation minus potential evapotranspiration) of vegetation period (March–November) RHmin % Minimum relative humidity during July and August a, b, c SWBmin§ mm Minimum site water balance (Grier & Running, 1977) a, b, c

*Variables were used for the genecological models of spruce (a), fir (b), and beech (c). †Values were calculated per year and then averaged across the past (1931–1960), current (1981–2000), and future (2021–2050, 2061–2090) time periods, if not otherwise stated. ‡Calculations were based on daily data, if not otherwise stated. §Calculations were based on monthly data. 5362 | FRANK ET AL. which is comparable to the new RCP6.0 scenario (Knutti & Sedlacek, mixed-effects model (lmer function in the “lme4” package; Bates, 2013). The best matching RCMs were used to downscale projections Machler,€ Bolker, & Walker, 2015). The model accounted for the from the global circulation model (GCM) ECHAM5 (Roeckner et al., fixed effect of early seedling height (covariate) and the random 2003), and represent a “dry”,an“intermediate”, and a “wet” future cli- effects of block, population, family-within-population, block by popu- mate within the A1B scenario (Remund et al., 2014). The selected lation interaction, and residual error, i.e., block by family-within- RCMs were CLM from the Max Planck Institute for Meteorology population interaction (details described in Frank, Pluess, et al., 2017 (Keuler et al., 2009), and RCA and RegCM3 from the “ENSEMBLES” and Frank, Sperisen, et al., 2017). The covariate early seedling height project (Hewitt & Griggs, 2004; Van Der Linden & Mitchell, 2009). was used to account for potential maternal effects and because Comparing 2071–2100 vs. 1981–2000, the average temperature replication and randomization of populations and families were not anomalies according to these three RCMs are +4.2°C, +3.2°C, and used in the nursery. Early seedling height was recorded after the +2.9°C in the north, and +4.3°C, +4.0°C, and +3.2°C in the south third growing season for spruce and fir, and after the second grow- (Remund et al., 2014). Average precipitation anomalies are 13.7%, ing seasons for beech. Prior to ANOVA, outliers were identified 7.8%, and +4.6% in the north, and 26.0%, 15.3%, and 8.0% in using the same linear mixed-effects model without the covariate. the south (Remund et al., 2014). For all three RCMs, climate projec- Outliers, which are observations whose residuals exceeded three tions of daily and monthly mean T, Td, and P were further down- standard deviations, were removed from the final dataset. The per- scaled using the Change Factor Method with 1981–2000 as the centage of outliers per trait was 0.5%–1.2% in spruce, 0.2%–0.8% in reference period (Tabor & Williams, 2010). Thereby, two datasets fir, and 0.6%–1.2% in beech. We used the variance components were used: (1) the reference data consisting of monthly mean values from ANOVA to estimate quantitative genetic parameters. In particu- for 1981–2000 on a 250 m grid that were calculated using the same lar, we calculated within-population additive genetic variation 2 2 weather stations and interpolation methods as described above (i.e., (r a(p) = 3r f(p)), assuming that the open-pollinated progeny included Shepard’s Gravity), and (2) the projected daily and monthly climate full-sibs and perhaps selfs, in addition to true half-sibs (Campbell, data on a 25 km grid obtained from the three RCMs. Using dataset 2, 1979). We also calculated population differentiation (Qst)as 2 2 2 daily and monthly climate anomalies were calculated, i.e., differences r p/[r p + 2r a(p)] (Spitze, 1993). In addition, the mixed models between a period of interest (p) and the reference period (r). This was provided estimates of population effects (i.e., Best Linear Unbiased done for the projected periods 2021–2050 and 2061–2090. Temper- Predictions, BLUPs) that were used for genecological modeling. ature anomalies were expressed as differences (TpTr), but precipitation anomalies as percentages ([P P ]/P ) to prevent negative p r r 2.4.2 | Genecological models values. These anomalies were then interpolated directly to every seed source location and NFI sample plot (Shepard’s Gravity Interpolation, We used multivariate genecological models to describe the popula- described above), and added (multiplied in case of precipitation) to tion variation in H, BudBreak, GrowthCess, and LeafSen for each the corresponding 1981–2000 grid cell value from the reference data species. These genecological models were derived from multiple lin- set 1 (see above) to obtain the actual projected values. Using these ear regressions of BLUPs and past seed-source climate data from values, projections for RH, PET, and SWB were calculated for every 1931 to 1960, i.e., the period closest to the establishment time of seed source location and NFI sample plot as described above. We the populations. From a larger set of 114 variables considered by then used the daily estimates for T, P, and RH, and the monthly val- Frank, Pluess, et al. (2017) and Frank, Sperisen, et al. (2017), we ues for PET and SWB to derive 30-year averages for the climate vari- evaluated 12 temperature and water-related variables (Table 1) for ables shown in Table 1. Future climate changes were calculated as which projections for future climate at seed-source locations and the differences between mean projected variables of 2021–2050 or NFI sample plots were reliable. These 12 variables were chosen 2061–2090 and mean measured variables of 1981–2000. based on (1) availability of past, present, and future climate data for all seed-source locations and NFI sample plots, (2) similarity to variables used in other genecological studies to facilitate comparisons, 2.4 | Data analyses (3) balance between temperature and precipitation variables (six All analyses were performed using the statistical computing environ- each), and (4) biological relevance. For example, we included some ment R (v3.2.4; R Core Team, 2016). Spatial calculations for the NFI monthly temperature variables (MTwarm and MTcold) because tem- sample plots, i.e., spatial modeling of population phenotypes and perature extremes are known to be important, but used water bal- estimating risks of maladaptation from climate change and seed ance of the complete vegetative period (March–November) to transfer (see below), were done using raster datasets and the R account for regionally different seasonalities in regards to precipita- packages “raster”, “maptools”, and “ncdf4”. tion and evapotranspiration. Using correlations among all 12 variables, we excluded two variables (MAT and PRCPETveg) that were very highly correlated (|r| ≥ .98) with the other variables. From the 2.4.1 | Variance components resulting ten variables, we chose eight subsets of variables that had Variance components for H, BudBreak, GrowthCess, and LeafSen low collinearity (VIF < 10; Dormann et al., 2013). Each of these sub- were derived from analysis of variance (ANOVA) using a linear sets contained one of four highly correlated temperature variables FRANK ET AL. | 5363

(MTsp, MTwarm, MTcold, CD; |r| ≥ .90), one of two combinations of from climate change. The transfer of FRM (i.e., “seed transfer”) is not precipitation variables (PRCan vs. PRCsu and PRCwi; |r| = .83–.93), explicitly regulated in Switzerland, but current practice follows regula- and two variables with lower correlations between each other and tions for mixing seedlots. Seeds are only allowed to be mixed when all other variables (CONT, RHmin, SWBmin; |r| ≤ .81). The linear and they are derived from the same forest region, i.e., Jura mountains, quadratic terms for the resulting eight subsets of climate variables Central Plateau, northern Alps, central Alps, and southern Alps, and were tested for each species and trait in multiple linear regressions from an elevational band of 200 m for stands located below using the all-subsets variable reduction approach (R function regsub- 1200 m a.s.l., or 100 m for stands located above 1200 m a.s.l. In sets, package “leaps”) and Mallows’ Cp selection criterion (Mallows, this study, we further distinguished between the western and eastern 1973). We chose the variable combinations that resulted in the best parts of the central Alps, as they often show distinct patterns of biodi- regression models per species judged by the traits’ average adjusted versity. The resulting regions represent the six main biogeographic R2 values. Model P values were corrected for multiple comparison regions of Switzerland (Figure 1; Gonseth et al., 2001). We calculated among traits (Bonferroni, n = number of traits per species, i.e., relative risk from seed transfer for every NFI sample plot of each spe- (three). The final variable subset included MTwarm, CONT, PRCsu, cies using past predicted population effects. Subsequently, we derived PRCwi, RHmin, and SWBmin for spruce, and MTsp, CONT, PRCsu, averages for each biogeographic region and elevation class (500 m). PRCwi, RHmin, and SWBmin for fir and beech (Table 1). 3 | RESULTS 2.4.3 | Modeled population phenotypes 3.1 | Quantitative genetic variation We predicted population phenotypes for all NFI sample plots from past (1931–1960), current (1981–2000), and future (2021–2050 and The traits assessed in this study showed considerable within-popula- 2 2061–2090) climates using the multivariate genecological models tion additive genetic variance (r a(p)) and population differentiation described above. Phenotypes were predicted separately for spruce, (Qst), and these values varied greatly among species and traits fir, and beech, i.e., for all NFI sample plots where each species cur- (Table 2). Population differentiation was clearly largest for seedling rently occurs (WSL, 2014). height of spruce (H; Qst = 0.48), followed by the phenological traits

of beech (BudBreak and LeafSen; Qst = 0.26 and 0.27). For fir and the remaining traits of spruce and beech, Q ranged between 0.10 2.4.4 | Risk of maladaptation from climate change st and 0.22. We used the relative risk index to estimate maladaptation due to climate change (Campbell, 1986; St. Clair & Howe, 2007). We calcu- 3.2 | Trait–climate associations lated two risk components, one describing current risk (CurrRisk), i.e., risk associated with differences in climate between 1931–1960 All seedling traits were significantly related to past seed source cli- and 1981–2000, and one describing future risk (FutRisk), i.e., risk mates as shown by the multivariate genecological regression models 2 associated with differences in climate between 1981–2000 and (PBonf < .05; Table 3). The highest model R adj for single traits was 2021–2050 (FutRisk1) and 2061–2090 (FutRisk2). For FutRisk, we obtained for H of spruce (.68), followed by leaf senescence of beech used climate projections from the three RCMs described above. Rel- (LeafSen; .47), and H of fir (.46). Spruce traits were predominantly ative risk was calculated for each species and NFI sample plot where associated with seed source temperature, whereas fir and beech spruce, fir, or beech were present as the proportion of nonoverlap traits were associated both with seed source temperature and water between two normal distributions centered at the predicted pheno- availability (Table 3). types for each climate period. That is, the predicted population effects defined the means of the normal distributions, and the 2 3.3 | Risks of maladaptation from climate change within-population additive genetic variation (r a(p)) defined the common variance (Appendix S2). Relative risk values range from 0 to 1, 3.3.1 | Current risk with 0 indicating “no risk” and 1 “high risk” of genetic maladaptation. We mapped CurrRisk and FutRisk for each species across Switzer- Current relative risk (CurrRisk), i.e., risk from recent past climate land, and derived mean values of risk for each species, trait, eleva- change between 1931–1960 and 1981–2000, averaged 0.07–0.26 tion class (500 m), and biogeographic region (Figure 1; Gonseth, for each species (Figure 2). Compared to average risks from seed Wohlgemuth, Sansonnens, & Buttler, 2001). transfer based on current Swiss guidelines (0.07–0.13; TransRisk), CurrRisk in spruce was consistently lower than TransRisk for all three traits (0.05–0.11). However, CurrRisk was larger than TransRisk 2.4.5 | Risk of maladaptation from seed transfer in most traits of fir and beech. Thereby, larger values than 0.25 were We also calculated relative risk of seed transfer using current Swiss found for H of fir (0.26) and LeafSen of beech (0.32). Beech was the regulations on the use of forest reproductive material (FRM) (EDI, species showing largest CurrRisk values (0.23–0.32) exceeding Trans- 1994). This risk served as a benchmark for evaluating relative risk Risk in magnitude and variation. 5364 | FRANK ET AL.

TABLE 2 Quantitative genetic statistics used to calculate risk of particularly for H. FutRisk2 of H was high across all of Switzerland genetic maladaptation due to climate change under all three RCMs, and in all biogeographic regions and elevational Within-population Population classes (Figures 3a and S2a–S7a in Appendix S1). High future risks additive genetic differentiation were also found for BudBreak and GrowthCess of spruce at low ele- † r2 Q Species Trait* Unit variation ( a(p)) ( st) vations (≤1000 m a.s.l.), and for GrowthCess in the uppermost eleva- Spruce H mm 83.10 0.48 tion class (2000–2500 m a.s.l.; Figs. S5a–S7a in Appendix S1). BudBreak JD 51.64 0.10 However, the latter result is based on only 90 spruce forest plots, GrowthCess JD 22.36 0.15 compared to 1832 spruce plots between 1500 and 2000 m a.s.l. Fir H mm 17.68 0.22 For fir, future risks for each trait averaged across all models were BudBreak JD 11.75 0.11 generally low, ranging between 0.04 and 0.35 for FutRisk1, and GrowthCess JD 0.73 0.17 between 0.13 and 0.26 for FutRisk2 (Figure 2b). An exceptionally Beech H mm 1571.76 0.19 high FutRisk1 value was found for H under the climate model CLM BudBreak JD 3.85 0.26 (0.52), which was almost five times larger than TransRisk. FutRisk2 of H was clearly higher in the southern Alps than in all other regions, LeafSen JD 10.74 0.27 for the CLM and RegCM3 models being greater than 0.60 (Fig- *H, total seedling height; BudBreak, timing of bud break; GrowthCess, ures 3b and S2b and S4b in Appendix S1). timing of height growth cessation; LeafSen, timing of leaf senescence, based on leaf coloration. For beech, future risks also increased with time for BudBreak and †JD is Julian Day, i.e., day of the year. LeafSen (FutRisk1 of 0.23–0.33 and 0.31–0.44 vs. FutRisk2 of 0.45– 0.58 and 0.40–0.54 per trait; Figure 2c). In particular, FutRisk2 of 3.3.2 | Future risk BudBreak using the CLM model was high at low elevations in the central and southern Alps, but also high in many parts of western and Future relative risks generally increased with time for spruce and northern Switzerland (Figure 3c). For H, in contrast, future risks dif- beech, with risk by the end of the century (FutRisk2) exceeding risk fered little from CurrRisk (0.14–0.22 vs. 0.16–0.31 per trait; Figure 2c). by mid-century (FutRisk1) in most traits (Figure 2). However, future Beech showed high variation in future risks, both within and among risks remained constantly low for fir. Relative risks associated with regions and elevation classes (Figures 3c, S2c–S7c in Appendix S1). future climates were generally larger for the regional climate model (RCM) CLM compared to the RCA and RegCM3 models. For spruce, future risks increased with time for all three traits. 4 | DISCUSSION Averaged over all three RCMs, FutRisk1 was 0.10–0.21 for each trait. These values were as low as TransRisk and CurrRisk, but mean We studied genetic maladaptation to future climates of spruce, fir, FutRisk2 per trait was five to eight times larger than TransRisk (0.33– and beech in Switzerland to identify species and regions that are 0.64; Figure 2a). Regional variation in future risk was generally low, most vulnerable to climate change. We used relative risk of genetic

TABLE 3 Genecological models, i.e., regression equations to predict population effects from climatic variables for key phenotypic traits of Norway spruce (P. abies), silver fir (A. alba), and European beech (F. sylvatica) seedlings. Significant regression coefficients are indicated in bold

(PBonf < .05). Climate variable abbreviations are explained in Table 1.

2 Species Traits R adj PBonf Genecological models Spruce H .68 <.001 Y = 88.40 + 4.425 MTwarm + 0.007 CONT2 + 0.032 SWBmin BudBreak .21 <.001 Y = 67.82 + 10.069 MTwarm 0.334 MTwarm2 0.002 CONT2 GrowthCess .36 <.001 Y = 64.22 + 7.938 MTwarm 0.243 MTwarm2 + 1.0E-05 PRCwi2 Fir H .46 <.001 Y = 146.33 + 3.898 CONT 0.029 CONT2 2.3E-05 PRCsu2 + 0.037 PRCwi + 0.005 RHmin2 + 7.5E-05 SWBmin2 BudBreak .24 <.001 Y = 55.11 0.120 CONT + 0.013 PRCsu + 0.040 PRCwi 9.1E-05 PRCwi2 2.186 RHmin + 0.021 RHmin2 GrowthCess .14 .002 Y = 0.71 + 0.099 MTsp + 1.4E-05 SWBmin2 Beech H .27 <.001 Y = 1115.11 0.557 PRCsu + 0.001 PRCsu2 0.001 PRCwi2 + 47.572 RHmin 0.454 RHmin2 + 0.001 SWBmin2 BudBreak .34 <.001 Y = 63.80 + 0.750 MTsp + 0.097 CONT + 2.119 RHmin 0.021 RHmin2 + 8.5E-05 SWBmin2 LeafSen .47 <.001 Y = 167.46 + 9.328 MTsp 0.503 MTsp2 6.305 CONT + 0.054 CONT2 0.112 PRCwi + 2.3E-04 PRCwi2 0.201 RHmin FRANK ET AL. | 5365

FIGURE 2 Relative risks of genetic maladaptation to climates for Norway spruce (a; P. abies), silver fir (b; A. alba), and European beech (c and d; F. sylvatica). Traits include seedling height (H) and bud break (BudBreak) for all three species, growth cessation (GrowthCess) for spruce and fir, and leaf senescence (LeafSen) for beech. Bars represent mean relative risks (SD) from average seed transfer (TransRisk), from past climate change between 1931–1960 and 1981–2000 (CurrRisk), and from future climate change between 1981–2000 and 2021–2050 (FutRisk1), and between 1981–2000 and 2061–2090 (FutRisk2). Past and current climates are based on measured historic data; future climates are based on the IPCC A1B scenario, general circulation model ECHAM5, and three regional climate models (CLM, RCA, and RegCM3) maladaptation, which is a function of the difference between the models is probably affected by unavoidable extrapolation, we do population phenotype for an adaptive trait in the climate in which it not expect major changes in the trends. More details on the evolved and the value of that trait that is expected to be adapted to uncertainty associated with model extrapolation are provided in a different climate, as well as the amount of within-population Fig. S8 (Appendix S1). genetic variance. To judge the degree of future maladaptation, we Our genecological models explained as much as 68% of pheno- used risks estimated from current practices of moving populations typic variation in spruce, 46% in fir, and 47% in beech. These num- for reforestation as comparison. bers are within the range of those found in comparable genecological studies, e.g., R2 of up to .58 in Douglas-fir (St. Clair & Howe, 2007) and .63 in black spruce (Picea mariana; Beaulieu, Per- 4.1 | Model performance ron, & Bousquet, 2004). R2 values for genecological models are The basis of risk estimation is the prediction of population means constrained by within-population sampling errors, population sam- for future climates based on models developed from past climates. pling errors, errors associated with the climate interpolation, and This approach involves two sources of uncertainty. Uncertainty variation in natural selection. For example, genecological models are increases with longer-term projections, because of the uncertainty typically used in a way that assumes populations are optimally in the climate projections and uncertainty in the extrapolation of adapted to their local environments, which might not be strictly genecological models to new climatic conditions. The climate data true (e.g., Rehfeldt, Ying, Spittlehouse, & Hamilton, 1999). Consider- used to develop the genecological models covered many of the ing these various sources of potential errors, our model perfor- future climate scenarios, but not all. In general, extrapolation was mance was reasonable, allowing us to draw valuable conclusions greater for the temperature-related variables, for fir and beech, from genecological relationships and risk estimates across time, spe- and for the 2061–2090 scenarios. Although the accuracy of the cies, and regions. 5366 | FRANK ET AL.

FIGURE 3 Geographic variation in relative risk of genetic maladaptation to the climates of 2061–2090 (FutRisk2). Risks are shown for seedling height (H) of Norway spruce (a; P. abies) and silver fir (b; A. alba), and for bud break (BudBreak) of European beech (c; F. sylvatica). FutRisk2 is based on the IPCC A1B scenario, the general circulation model ECHAM5, and the regional climate model CLM. The six main biogeographic regions of Switzerland (Figure 1) are indicated by their boundaries

Therefore, CurrRisk represents an acceptable level of risk in all three 4.2 | Maladaptation from seed transfer and current species. climate change

As a benchmark to judge the degree of genetic maladaptation, we 4.3 | Maladaptation to climate increases with time used risks associated with current practices of moving populations for reforestation. Average seed transfer risks for spruce, fir, and Our results indicate that current populations of spruce, fir, and beech in Switzerland (TransRisk) were as high as 0.18 per trait (Fig- beech are sufficiently adapted to the projected climates of 2021 to ure 2), which is similar to transfer risks in Douglas-fir, ponderosa 2050, with FutRisk1 being similar to TransRisk and CurrRisk (Fig- pine (Pinus ponderosa), and sugar pine (Pinus lambertiana) in the Paci- ure 2). The exceptionally high value of FutRisk1 observed for seed- fic Northwest (~0.2–0.3; Campbell & Sugano, 1987; Sorensen, 1994; ling height of fir under the climate model CLM can be explained by St. Clair & Howe, 2007). Thus, TransRisk associated with current the larger decrease in winter precipitation projected by this model practices is a valuable benchmark for evaluating maladaptation to between 2021–2050 and 2061–2090 (Fig. S9e, Appendix S1). In future climates. general, uncertainties in climate projections are larger for precipita- Current risks from climate change between 1931–1960 and tion than for temperature (CH2011, 2011), limiting firm conclusions 1981–2000 (CurrRisk) were as low as TransRisk in spruce (Figure 2). regarding the impact of precipitation changes. By 2061–2090, our In fir and beech, CurrRisk was higher than TransRisk, indicating that results suggest that risk of maladaptation will remain low for fir, but these species may already be experiencing adaptational lag (Maty as, will increase markedly for spruce and beech, with similar trends 1990). Nevertheless, even the largest value of CurrRisk for leaf associated with all three RCMs (Figure 2). Consequently, spruce and senescence in beech (0.32) was comparable to accepted transfer beech will likely suffer from significant maladaptation by the end of risks in other species, e.g., 0.3 in ponderosa pine (Sorensen, 1994). the century, but probably not fir. FRANK ET AL. | 5367

Therefore, beech stands in these areas need special attention by for- 4.4 | Species-specific patterns of maladaptation est managers. Although risk was generally low for fir, vulnerability reflect adaptive strategies was higher in the southern Alps based on FutRisk2 for height (Fig- How can we explain differences in projected maladaptation ure 3b). This effect is associated with the particularly large decrease between spruce, fir, and beech? Our results show that all three spe- in summer precipitation in the southern Alps by 2061–2090 cies exhibit large within-population genetic variance, similar to the (Fig. S11, Appendix S1). variance found in Douglas-fir (St. Clair & Howe, 2007). In fact, most forest trees show large amounts of within-population genetic varia- 4.6 | Potential consequences of maladaptation to tion (Alberto et al., 2013; Howe et al., 2003). High levels of genetic future climates variation facilitate in situ evolutionary adaptation of populations, and lower relative risks of maladaptation from climate change by Our data provide information about the genetic difference between reducing the degree of nonoverlap between current populations vs. current population phenotypes and the phenotypes that are pro- populations expected to be well adapted to future climates (St. Clair jected for these populations to be well-adapted in the future. Since & Howe, 2007). The projected amounts of future climate change risk is a relative value for which no empirical data for phenotypic were also similar across the current distributions of these three performance are available, we cannot predict the actual conse- species in Switzerland (Fig. S10, Appendix S1). Our results agree quences on tree fitness associated with particular relative risk val- with general climate trends showing slight warming and drying until ues. Yet, risk values above a threshold of 0.3 are generally 2050, and larger increases in temperature and drought-related cli- considered to be biologically relevant, because they exceed the mate variables until the end of the century (Fig. S9, Appendix S1; amount of risk that is associated with current seed transfer prac- CH2011, 2011). tices (see above). Our genecological models indicate that potential However, population differentiation and trait-climate associations consequences are related to trade-offs involving the synchronization clearly differed among the species, being largest for spruce, moder- of populations’ annual growth cycles with their local temperature ate for beech, and rather low for fir (Tables 2 and 3, see also Frank, regimes. For example, if seedlings grow too late into the fall, they Pluess, et al., 2017 and Frank, Sperisen, et al., 2017). These contrast- will be susceptible to frost injury and mortality. Yet, if they stop ing genecological patterns represent different adaptive strategies growth too early in the season, the seedlings may show lower com- (Frank, Sperisen, et al., 2017; Rehfeldt, 1994), leading to differences petitive ability (Howe et al., 2003). In addition, trade-offs involving in projected future maladaptation among species. Risks were highest water availability will likely contribute to maladaptation in the for spruce, which is under strong selection by local temperature future. Therefore, moving away from the phenotypic optimum for a regimes (Table 3). Consequently, future maladaptation in this adap- population in either direction is likely to be biologically relevant. tive specialist will be driven mainly by climate warming. Considerable For spruce, the greatest risk by the end of the century was levels of future climatic maladaptation were also found for beech, found for seedling height. Height growth integrates multiple traits which is associated with both local temperature and water availabil- related to growth and phenology, as well as predisposition to frost ity. Therefore, maladaptation in this species will be determined and drought hardiness. Furthermore, it is often used as a surrogate largely by a combination of these climate variables. Fir was classified for plant fitness (Kapeller, Lexer, Geburek, Hiebl, & Schueler, as an adaptive generalist, with rather low population differentiation 2012). Our genecological models indicate that spruce populations and weaker relationships between seedling traits and seed source with greater height growth will be better adapted in a warming cli- climates. Thus, fir is less likely to become maladapted to future mate (Fig. S12, Appendix S1). Thus, the risk of maladaptation may climates than spruce and beech. be associated with diminished competitive ability. Because minimum site water balance was also included in the genecological model for seedling height, this indicates that increased drought is 4.5 | Regional variation in maladaptation to future also associated with future maladaptation. Indeed, spruce is sensi- climates tive to drought stress, which results in less diameter growth in Future risks for spruce were generally high across Switzerland (Fig- warm and dry seasons (Lebourgeois et al., 2010; Zang, Hartl-Meier, ure 3). This is because temperature drives variation in spruce, and Dittmar, Rothe, & Menzel, 2014). Also, molecular genetic variation projected temperature increases were quite uniform across Switzer- in spruce in the south-eastern Alps was associated with seed land (Fig. S11, Appendix S1). In contrast, we found variation in mal- source precipitation variables indicating local adaptation to water adaptation among regions for beech and fir (Figure 3). Risks for bud availability (Di Pierro et al., 2016). In addition, the high risks asso- break, an important adaptive trait of beech (Frank, Sperisen, et al., ciated with spruce bud break and growth cessation at low eleva- 2017), were high for several regions. These include low-elevation tions match the results of previous studies. Site productivity of areas in southern, western, and northern Switzerland, and in the spruce in Central Europe, for example, is projected to decrease inner-Alpine valleys (Figure 3c). Several of those areas belong to the under climate change at low elevations, but increase at higher ele- currently driest regions of Switzerland, which are projected to vations (Hlasny et al., 2011; Nothdurft et al., 2012). Consequently, become even drier in the future (Remund & Augustin, 2015). spruce is expected to lose large fractions of its current range in 5368 | FRANK ET AL. the lowlands, and expand to higher elevations in the mountains. functions that predict how the growth of lodgepole pine and Scots At intermediate elevations, however, it will likely experience pine (Pinus sylvestris) will change in response to climate change indi- enhanced competition from beech, which is also projected to cate that evolutionary adaptation will probably be insufficient to expand its range upwards (e.g., Hlasny et al., 2011). This could fur- avoid future maladaptation (Rehfeldt, Wykoff, & Ying, 2001; Rehfeldt ther narrow the future range of spruce. et al., 2002). Spruce, fir, and beech all have high levels of among- For beech, we found high risks for bud break and leaf senes- population gene flow based on isozymes (spruce and fir; Finkeldey, cence, indicating that this species could suffer from phenological Maty as, Sperisen, & Bonfils, 2000) and nuclear microsatellites (beech; mismatch in the future. As temperatures increase, the high chil- Pluess et al., 2016). Our results show that modeled past and current ling requirements of beech for release of endodormancy may no population phenotypes vary at small scales, which facilitates the longer be fulfilled (Murray, Cannell, & Smith, 1989). This may among-population exchange of preadapted alleles and enhances evo- lead to delayed bud break and reduced growth. In contrast, lutionary adaptation (Kremer et al., 2012). Nevertheless, preadapted increasing fall temperatures could also delay leaf senescence, alleles can be rare or geographically distant in Switzerland as indi- thereby prolonging the growing season and exposing beech to cated by the comparison of past and future modeled phenotypes higher risks of early frost damage (Lebourgeois et al., 2010; (seedling height and bud break) for regions at high risk of future mal- Vitasse, Porte, Kremer, Michalet, & Delzon, 2009). Yet, the adaptation (Fig. S12, Appendix S1). Finally, phenotypic plasticity and impacts of global warming on phenological timing of trees are epigenetic effects allow populations to acclimate to climate change still largely unknown and a matter of debate (Korner€ & Basler, (Alfaro et al., 2014; Nicotra et al., 2010; Park et al., 2014). However, 2010a, 2010b). In addition, our models suggest that genetic vari- the mere existence of high population differentiation in spruce and ation in beech phenology is not only related to temperature, but beech indicates that phenotypic plasticity and epigenetics alone will also to water availability. Consequently, it is difficult to draw be insufficient for coping with climate change. Thus, further strategies firm conclusions regarding the nature of future phenological mal- to introduce preadapted alleles to high-risk populations are required. adaptation in beech. Previous modeling indicates that at low elevations, beech site productivity will probably decrease under 4.8 | Adjust forest management practices to climate change (Hlasny et al., 2011; Nothdurft et al., 2012). promote climate change adaptation Because beech mostly occurs at low elevations (Bolte, Cza- jkowski, & Kompa, 2007), beech occurrence is projected to In practice, relative risk helps to infer which species and regions are decrease in these regions, shifting to higher elevations and more most vulnerable to climate change. Our results indicate that forest northern areas by the end of the century (Meier et al., 2011). management for spruce should be adjusted under climate change. For fir, our results suggest that maladaptation may occur in Current Swiss forestry depends largely on spruce as its “bread-and- southern Switzerland, due to projected decreases in summer precipi- butter tree”. It is the most abundant conifer in Switzerland, providing tation (see above). Tree ring analyses in southern Germany and Aus- highly valuable timber (Cioldi et al., 2010). Thereby, spruce timber tria have shown that fir generally exhibits higher drought resistance production is most profitable in the Swiss lowlands, where produc- and resilience than spruce and beech (Zang et al., 2014). Vegetation tivity is highest and harvesting costs are lowest. Therefore, these modeling indicates that fir can become codominant in areas where areas deserve the most attention. Climate change effects should also summer precipitation does not fall below 120–150 mm (Tinner et al., be considered for beech stands in several southern, northern, west- 2013). For the more extreme RCM, summer precipitation is pro- ern, and inner-Alpine (Valais) areas of Switzerland. These regions are jected to decrease in southern Switzerland to about 200–400 mm already particularly dry and prone to future droughts (Remund & by 2061–2090. However, this is likely sufficient for the persistence Augustin, 2015). of fir in this area. Several management strategies can be used to mitigate the negative effects of climate change on forests (Aitken & Whitlock, 2013; Brang et al., 2014; Lefevre et al., 2014; Schelhaas et al., 2015). The 4.7 | How can current populations avoid objective of most strategies is to enhance gene flow (or migration) maladaptation to future climates? and evolutionary adaptation. Ideally, different strategies are applied The scenarios outlined above assume that future populations remain and combined in a flexible manner (Brang et al., 2014). For species genetically identical to current populations. However, populations and regions at low risk of maladaptation, such as fir in northern undergo constant evolutionary processes, mainly driven by the bal- Switzerland, silvicultural strategies should aim at enhancing regenera- ance of selection and gene flow (Savolainen et al., 2007). Our results tion such that natural selection can continuously act on large num- project a large genetic mismatch for spruce and beech by the end of bers of juvenile trees (Kramer et al., 2008; Lefevre et al., 2014). For this century. That is, current populations must either evolve quickly species and regions at high risk of maladaptation to future climates, or show large plastic responses to avoid maladaptation. There is con- such as spruce in most parts of Switzerland and beech in drought- siderable within-population genetic variation in all three species, indi- prone regions, forest management strategies should consider assisted cating a high potential for in situ evolution (Alfaro et al., 2014; Frank, gene flow to reduce climate change vulnerability (Aitken & Whitlock, Pluess, et al., 2017; Frank, Sperisen, et al., 2017). However, response 2013; Williams & Dumroese, 2013). To this end, climate-based seed FRANK ET AL. | 5369 transfer guidelines are needed, preferably ones that largely ignore Alfaro, R. I., Fady, B., Vendramin, G. G., Dawson, I. K., Fleming, R. A., Saenz- ... administrative boundaries such as state or country borders. Romero, C., Loo, J. (2014). The role of forest genetic resources in responding to biotic and abiotic factors in the context of anthropogenic Finding suitable seed sources is a difficult task (Potter & Har- climate change. Forest Ecology and Management, 333,76–87. grove, 2012). Our results can contribute to developing seed transfer Bates, D., Machler,€ M., Bolker, B. M., & Walker, S. C. (2015). Fitting linear guidelines for Switzerland that take into account future climate mixed-effects models using lme4. Journal of Statistical Software, 67, change, but will need to be carefully evaluated in relation to current 1–48. Beaulieu, J., Perron, M., & Bousquet, J. (2004). Multivariate patterns of forest practices. First, we can try to identify regions with current adaptive genetic variation and seed source transfer in Picea mariana. phenotypes that are similar to those that would be adapted to Canadian Journal of Forest Research, 34, 531–545. regions with high projected maladaptation to future climates. Maps Begert, M., Seiz, G., Foppa, N., Schlegel, T., Appenzeller, C., & Muller,€ G. € showing past and future modeled population effects can be used for (2007). Die Uberfuhrung€ der klimatologischen Referenzstationen der Schweiz in das Swiss National Basic Climatological Network (Swiss that purpose (Fig. S12, Appendix S1). However, our results suggest NBCN). Arbeitsberichte der MeteoSchweiz, 215, 43. that suitable regions and stands are rare in Switzerland. Second, the Bolte, A., Czajkowski, T., & Kompa, T. (2007). The north-eastern distribu- strong temperature associations in spruce will allow us to develop tion range of European beech – a review. Forestry, 80, 413–429. elevational seed transfer guidelines for this species, facilitating the Bosela, M., Popa, I., Gom€ ory,€ D., Longauer, R., Tobin, B., Kyncl, J., ... € transfer or spruce from lower to higher elevations. For this strategy, Buntgen, U. (2016). Effects of post-glacial phylogeny and genetic diversity on the growth variability and climate sensitivity of European the higher frost susceptibility of low-elevation populations planted at silver fir. Journal of Ecology, 104, 716–724. higher elevations has to be considered. Also, probably no suitable Brang, P., Spathelf, P., Larsen, J. B., Bauhus, J., Boncına, A., Chauvin, C., seed sources are available for the populations at the lowest eleva- ... Svoboda, M. (2014). Suitability of close-to-nature silviculture for tions. These limitations of regional and elevational seed transfer adapting temperate European forests to climate change. Forestry, 87, 492–503. guidelines imply that climate change adjusted management recom- Bussotti, F., Pollastrini, M., Holland, V., & Bruggemann,€ W. (2015). Func- mendations should also include more drastic options, such as the tional traits and adaptive capacity of European forests to climate introduction of forest reproductive material (FRM) from potentially change. Environmental and Experimental Botany, 111,91–113. drought-adapted stands in southern or southeastern Europe for Campbell, R. K. (1979). Genecology of Douglas-fir in a watershed in the Oregon Cascades. Ecology, 60, 1036–1050. intermixture with Swiss stands on dry sites. Also, the transfer of Campbell, R. K. (1986). Mapped genetic variation of Douglas-fir to guide FRM from regions where the species possess higher genetic variabil- seed transfer in southwest Oregon. Silvae Genetica, 35,85–96. ity can be used as an additional strategy (e.g., Bosela et al., 2016). Campbell, R. K., & Sugano, A. I. (1987). Seed zones and breeding zones Finally, local introduction or promotion of substitute species should for sugar pine in southwestern Oregon. Res. Pap. PNW-RP-383. U.S. Department of Agriculture, Forest Service, Pacific Northwest be considered, such as Douglas-fir as replacement for spruce, or Research Station, Portland, OR, USA. 18 p. oaks (Quercus spp.) as replacement for beech. Even if uncertainty CH2011 (2011). Swiss climate change scenarios CH2011. Zurich, Switzer- about climate change is likely to remain large (Lindner et al., 2014), land: C2SM, MeteoSwiss, ETH, NCCR Climate, and OcCC. management decisions should be taken soon owing to the long time Cioldi, F, Baltensweiler, A., Brandli,€ U-B., Duc, P., Ginzler, C., Herold Bonardi, A., ... Ulmer, U. (2010). Waldressourcen. In U.-B. Br€andli needed to implement new forest management strategies and to con- (Ed.), Schweizerisches Landesforstinventar. Ergebnisse der dritten Erhe- vert highly vulnerable forests to less susceptible ecosystems (Schel- bung 2004–2006 (pp. 30–113). Birmensdorf and Bern, Switzerland: haas et al., 2015; Temperli, Bugmann, & Elkin, 2012). Eidgenossische€ Forschungsanstalt fur€ Wald, Schnee und Landschaft (WSL) und Bundesamt fur€ Umwelt (BAFU). Di Pierro, E. A., Mosca, E., Rocchini, D., Binelli, G., Neale, D. B., & La ACKNOWLEDGEMENTS Porta, N. (2016). Climate-related adaptive genetic variation and population structure in natural stands of Norway spruce in the south-east- We wish to thank Jan Remund, Nick Zimmermann, and Oliver ern Alps. Tree Genetics & Genomes, 12, 16. Jakoby for sharing their knowledge of climate data. In addition, we Dormann, C. F., Elith, J., Bacher, S., Buchmann, C., Carl, G., Carre, G., ... are grateful to Jonas Stillhard, Rafael Wuest,€ and Jan Wunder for Lautenbach, S. (2013). Collinearity: A review of methods to deal with it and a simulation study evaluating their performance. Ecography, 36, their technical assistance in R computation, and to Andrea R. Pluess, 27–46. € Jurgen Zell, and Rita Gosh for statistical advice. This study was DWD (1979) Deutscher Wetterdienst, Aspirations- und Psychrometertafeln, funded by the research program “Forests and Climate Change” of 6th edn. Wiesbaden, Germany: Friedrich Vieweg & Sohn, Braun- FOEN and WSL. schweig. EDI (1994). Verordnung uber€ forstliches Vermehrungsgut, Schweiz- erische Eidgenossenschaft, Departement des Innern. Retrieved from REFERENCES www.admin.ch/opc/de/classified-compilation/19940363/index.html (accessed 21 February 2017). Aitken, S. N., & Whitlock, M. C. (2013). Assisted gene flow to facilitate EUFORGEN (European forest genetic resources programme) (2009). local adaptation to climate change. Annual Review of Ecology, Evolu- Distribution maps of Norway spruce (Picea abies), silver fir (Abies tion, and Systematics, 44, 367–388. alba), and beech (Fagus sylvatica). Retreived from www.euforgen.org/ Alberto, F. J., Aitken, S. N., Alıa, R., Gonzalez-Mart ınez, S. C., H€anninen, distribution-maps (accessed 2 June 2016). H., Kremer, A., ... Savolainen, O. (2013). Potential for evolutionary Finkeldey, R., Matyas, G., Sperisen, C., & Bonfils, P. (2000). Strategien zur responses to climate change – evidence from tree populations. Global Auswahl forstlicher Genreservate in der Schweiz. Forest Snow Land- Change Biology, 19, 1645–1661. scape Research, 75, 137–152. 5370 | FRANK ET AL.

Frank, A., Pluess, A. R., Howe, G. T., Sperisen, C., & Heiri, C. (2017). Lebourgeois, F., Rathgeber, C. B. K., & Ulrich, E. (2010). Sensitivity of Quantitative genetic differentiation and phenotypic plasticity of Euro- French temperate coniferous forests to climate variability and pean beech in a heterogeneous landscape: Indications for past cli- extreme events (Abies alba, Picea abies and Pinus sylvestris). Journal of mate adaptation. Perspectives in Plant Ecology, Evolution and Vegetation Science, 21, 364–376. Systematics, 26,1–13. Lefevre, F., Boivin, T., Bontemps, A., Courbet, F., Davi, H., Durand- Frank, A., Sperisen, C., Howe, G. T., Brang, P., Walthert, L., St. Clair, J. B., Gillmann, M., ... Pichot, C. (2014). Considering evolutionary processes & Heiri, C. (2017). Distinct genecological patterns in seedlings of in adaptive forestry. Annals of Forest Science, 71,723–739. Norway spruce and silver fir from a mountainous landscape. Ecology, Lindner, M., Fitzgerald, J. B., Zimmermann, N. E., Reyer, C., Delzon, S., 98, 211–227. Van Der Maaten E., ... Hanewinkel, M. (2014). Climate change and Gessler, A., Keitel, C., Kreuzwieser, J., Matyssek, R., Seiler, W., & Rennen- European forests: What do we know, what are the uncertainties, and berg, H. (2007). Potential risks for European beech (Fagus sylvatica L.) what are the implications for forest management? Journal of Environ- in a changing climate. Trees, 21,1–11. mental Management, 146,69–83. Gonseth, Y., Wohlgemuth, T., Sansonnens, B., & Buttler, A. (2001). Die Lindner, M., Maroschek, M., Netherer, S., Kremer, A., Barbati, A., Garcia- biogeographischen Regionen der Schweiz. Erlauterungen€ und Einteilungs- Gonzalo, J., ... Marchetti, M. (2010). Climate change impacts, adap- standard. Bern, Switzerland: Bundesamt fur€ Umwelt, Wald und Land- tive capacity, and vulnerability of European forest ecosystems. Forest schaft (BUWAL). Ecology and Management, 259, 698–709. Grier, C. G., & Running, S. W. (1977). Leaf area of mature northwestern Mallows, C. L. (1973). Some comments on Cp. Technometrics, 15, 661– coniferous forests: Relation to site water balance. Ecology, 58, 893– 675. 899. Matyas, C. (1990). Adaptation lag: A general feature of natural popula- Hanewinkel, M., Cullmann, D. A., Schelhaas, M. J., Nabuurs, G. J., & Zim- tions. In: Adaptability of seed sources across geographic zones. Proc. mermann, N. E. (2013). Climate change may cause severe loss in the WFGA-IUFRO symp. on breeding and genetic resources of conifers, economic value of European forest land. Nature Climate Change, 3, paper 2.226. Olympia, WA, USA. 10 p. 203–207. Meier, E. S., Edwards, T. C., Kienast, F., Dobbertin, M., & Zimmermann, Hewitt, C. D., & Griggs, D. J. (2004). Ensembles-based predictions of cli- N. E. (2011). Co-occurrence patterns of trees along macro-climatic mate changes and their impacts. Eos, 85, 566. gradients and their potential influence on the present and future dis- Hlasny, T., Barcza, Z., Fabrika, M., Balazs, B., Churkina, G., Pajtık, J., ... tribution of Fagus sylvatica L. Journal of Biogeography, 38, 371–382. Turcani, M. (2011). Climate change impacts on growth and carbon Modrzynski, J. (2007). Outline of Ecology. In M. G. Tjoelker, A. Boratyn- balance of forests in Central Europe. Climate Research, 47, 219–236. ski, & W. Bugala (Eds.), Biology and Ecology of Norway Spruce (pp. Houston Durrant, T., De Rigo, D, & Caudullo, G. (2016). Fagus sylvatica 195–253). Dordrecht, The Netherlands: Springer. and other beeches in Europe: Distribution, habitat, usage and threats. Murray, M. B., Cannell, M. G. R., & Smith, R. I. (1989). Date of budburst In J. San-Miguel-Ayanz, D. De Rigo, G. Caudullo, T. Houston Durrant of fifteen tree species in Britain following climatic warming. Journal & A. Mauri (Eds.), European atlas of forest tree species (pp. 94–95). of Applied Ecology, 26, 693–700. Luxembourg: Publ. Off. EU. Nakicenovic, N., & Swart, R. (2000). IPCC special report on emissions sce- Howe, G. T., Aitken, S. N., Neale, D. B., Jermstad, K. D., Wheeler, N. C., narios. Cambridge, UK: Intergovernmental Panel on Climate Change & Chen, T. H. H. (2003). From genotype to phenotype: Unraveling (IPCC). the complexities of cold adaptation in forest trees. Canadian Journal Nicotra, A. B., Atkin, O. K., Bonser, S. P., Davidson, A. M., Finnegan, E. J., of Botany, 81, 1247–1266. Mathesius, U., ... Van Kleunen, M. (2010). Plant phenotypic plasticity Kapeller, S., Lexer, M. J., Geburek, T., Hiebl, J., & Schueler, S. (2012). in a changing climate. Trends in Plant Science, 15, 684–692. Intraspecific variation in climate response of Norway spruce in the Nothdurft, A. (2013). Spatio-temporal prediction of tree mortality based eastern Alpine range: Selecting appropriate provenances for future on long-term sample plots, climate change scenarios and parametric climate. Forest Ecology and Management, 271,46–57. frailty modeling. Forest Ecology and Management, 291,43–54. Keuler, K., Lautenschlager, M., Wunram, C., Keup-Thiel, E., Schubert, M., Nothdurft, A., Wolf, T., Ringeler, A., Bohner,€ J., & Saborowski, J. (2012). Will, A., ... Boehm, U. (2009). Climate simulation with CLM, scenario Spatio-temporal prediction of site index based on forest inventories A1B run no.2, data stream 2: European region MPI-M/MaD. Hamburg: and climate change scenarios. Forest Ecology and Management, 279, World Data Center for Climate (WDCC). 97–111. Knutti, R., & Sedlacek, J. (2013). Robustness and uncertainties in the new Park, A., Puettmann, K., Wilson, E., Messier, C., Kames, S., & Dhar, A. CMIP5 climate model projections. Nature Climate Change, 3, 369–373. (2014). Can boreal and temperate forest management be adapted to Korner,€ C., & Basler, D. (2010a). Phenology under global warming. the uncertainties of 21st century climate change? Critical Reviews in Science, 327, 1461–1462. Plant Sciences, 33, 251–285. Korner,€ C., & Basler, D. (2010b). Warming, Photoperiods, and Tree Pepin, N., Bradley, R. S., Diaz, H. F., Baraer, M., Caceres, E. B., Forsythe, Phenology Response. Science, 329, 278–278. N., ... Yang, D. Q. (2015). Elevation-dependent warming in mountain Kramer, K., Buiteveld, J., Forstreuter, M., Geburek, T., Leonardi, S., regions of the world. Nature Climate Change, 5, 424–430. Menozzi, P., ... Van Der Werf, D. C. (2008). Bridging the gap Pluess, A. R., Frank, A., Heiri, C., Lalague,€ H., Vendramin, G. G., & between ecophysiological and genetic knowledge to assess the adap- Oddou-Muratorio, S. (2016). Genome-environment association study tive potential of European beech. Ecological Modelling, 216, 333–353. suggests local adaptation to climate at the regional scale in Fagus Kremer, A., Ronce, O., Robledo-Arnuncio, J. J., Guillaume, F., Bohrer, G., sylvatica. New Phytologist, 210, 589–601. Nathan, R., ... Schueler, S. (2012). Long-distance gene flow and Potter, K. M., & Hargrove, W. W. (2012). Determining suitable locations adaptation of forest trees to rapid climate change. Ecology Letters, 15, for seed transfer under climate change: A global quantitative method. 378–392. New Forests, 43, 581–599. Langlet, O. (1971). Two hundred years genecology. Taxon, 20, 653–722. R Core Team (2016). R: A language and environment for statistical comput- Lebourgeois, F., Pierrat, J. C., Perez, V., Piedallu, C., Cecchini, S., & Ulrich, ing. Vienna, Austria: R Foundation for Statistical Computing. E. (2010). Simulating phenological shifts in French temperate forests Rehfeldt, G. E. (1994). Evolutionary genetics, the biological species, and under two climatic change scenarios and four driving global circula- the ecology of the interior cedar-hemlock forests. In D. M. Baumgart- tion models. International Journal of Biometeorology, 54, 563–581. ner, J. E. Lotan, & J. R. Tonn (Eds.), Proceedings of the interior cedar- FRANK ET AL. | 5371

hemlock-white pine forests: Ecology and management, Spokane, WA (pp. Teepe, R., Dilling, H., & Beese, F. (2003). Estimating water retention 91–100). Pullman, WA, USA: Washington State University Extension. curves of forest soils from soil texture and bulk density. Journal of Rehfeldt, G. E., Tchebakova, N. M., Parfenova, Y. I., Wykoff, W. R., Kuz- Plant Nutrition and Soil Science, 166, 111–119. mina, N. A., & Milyutin, L. I. (2002). Intraspecific responses to climate Temperli, C., Bugmann, H., & Elkin, C. (2012). Adaptive management for in Pinus sylvestris. Global Change Biology, 8, 912–929. competing forest goods and services under climate change. Ecological Rehfeldt, G. E., Wykoff, W. R., & Ying, C. C. (2001). Physiologic plasticity, Applications, 22, 2065–2077. evolution, and impacts of a changing climate on Pinus contorta. Cli- Tinner, W., Colombaroli, D., Heiri, O., Henne, P. D., Steinacher, M., Unte- matic Change, 50, 355–376. necker, J., ... Valsecchi, V. (2013). The past ecology of Abies alba Rehfeldt, G. E., Ying, C. C., Spittlehouse, D. L., & Hamilton, D. A. (1999). provides new perspectives on future responses of silver fir forests to Genetic responses to climate in Pinus contorta: Niche breadth, climate global warming. Ecological Monographs, 83, 419–439. change, and reforestation. Ecological Monographs, 69, 375–407. Van Der Linden, P., & Mitchell, J. F. B. (2009). ENSEMBLES: Climate Remund, J. (2016). Meteodaten fur€ Forschungsprojekt WSL. Langjahrige€ change and its impacts: Summary of research and results from the Zeitreihen und Analyse basierend auf gleichbleibenden Wetterstationen. ENSEMBLES project. Exeter, UK: Met Office Hadley Centre. Bern, Switzerland: Meteotest. Vitasse, Y., Porte, A. J., Kremer, A., Michalet, R., & Delzon, S. (2009). Remund, J., & Augustin, S. (2015). Zustand und Entwicklung der Trocken- Responses of canopy duration to temperature changes in four tem- heit in Schweizer Waldern.€ Schweizerische Zeitschrift fur€ Forstwesen, perate tree species: Relative contributions of spring and autumn leaf 166, 352–360. phenology. Oecologia, 161, 187–198. Remund, J., Frehner, M., Walthert, L., K€agi, M., & Rihm, B. (2011). Williams, M. I., & Dumroese, R. K. (2013). Preparing for climate change: Schatzung€ standortspezifischer Trockenstressrisiken in Schweizer Forestry and assisted migration. Journal of Forestry, 111, 287–297. Waldern.€ Schlussbericht Version 2.3. Bern, Switzerland: Meteotest. Wolf, H. (2003). EUFORGEN Technical guidelines for genetic conservation Remund, J., Rihm, B., & Huguenin-Landl, B. (2014). Klimadaten fur€ die and use for silver fir (Abies alba). Rome, Italy: International Plant Waldmodellierung fur€ das 20. und 21. Jahrhundert. Bern, Switzerland: Genetic Resources Institute. Meteotest. WSL (2014). Swiss National Forest Inventory (NFI). Data of the third sur- Roeckner, E., B€auml, G., Bonaventura, L., Brokopf, R., Esch, M., Giorgetta, vey 2004/06 (NFI3). Fabrizio Cioldi 08.12.2014. Birmensdorf, Swit- M., ... Tompkins, A. (2003). The atmospheric general circulation model zerland: Swiss Federal Institute for Forest, Snow and Landscape ECHAM5, part 1, model description. Hamburg, Germany: Max-Planck- Research WSL. Institut fur€ Meteorologie. Zang, C., Hartl-Meier, C., Dittmar, C., Rothe, A., & Menzel, A. (2014). Pat- Romanenko, V. A. (1961). Computation of the autumn soil moisture using terns of drought tolerance in major European temperate forest trees: a universal relationship for a large area. In Proceedings of the Ukrai- Climatic drivers and levels of variability. Global Change Biology, 20, nian Hydrometeorological Research Institute, No. 3. Kiev, Ukraine. 3767–3779. Savolainen, O., Pyh€aj€arvi, T., & Knurr,€ T. (2007). Gene flow and local Zelenka, A., Czeplak, G., D’Agostino, V., Josefsson, W., Maxwell, E., Perez, adaptation in trees. Annual Review of Ecology, Evolution, and Systemat- R., ... Festa, R. (1992). Techniques for supplementing solar radiation ics, 38, 595–619. network data. Paris, France: IEA. Schelhaas, M. J., Nabuurs, G. J., Hengeveld, G., Reyer, C., Hanewinkel, M., Zimmermann, N. E., & Cullmann, D. (2015). Alternative forest management strategies to account for climate change-induced productivity and species suitability changes in Europe. Regional Environ- SUPPORTING INFORMATION mental Change, 15, 1581–1594. Additional Supporting Information may be found online in the sup- Sorensen, F. C. (1994). Genetic variation and seed transfer guidelines for ponderosa pine in central Oregon. Res. Pap. PNW-RP-472. U.S. porting information tab for this article. Department of Agriculture, Forest Service, Pacific Northwest Research Station, Portland, OR, USA. 24 p. Spitze, K. (1993). Population structure in Daphnia obtusa: Quantitative genetic and allozymic variation. Genetics, 135, 367–374. How to cite this article: Frank A, Howe GT, Sperisen C, et al. St. Clair, J. B., & Howe, G. T. (2007). Genetic maladaptation of coastal Risk of genetic maladaptation due to climate change in three Douglas-fir seedlings to future climates. Global Change Biology, 13, major European tree species. Glob Change Biol. 1441–1454. 2017;23:5358–5371. https://doi.org/10.1111/gcb.13802 St. Clair, J. B., Mandel, N. L., & Vance-Borland, K. W. (2005). Genecology of Douglas-fir in western Oregon and Washington. Annals of Botany, 96, 1199–1214. Tabor, K., & Williams, J. W. (2010). Globally downscaled climate projections for assessing the conservation impacts of climate change. Eco- logical Applications, 20, 554–565.