Stable Allometric Trajectories in Picea abies (L.) Karst. Trees along an Elevational Gradient

The effect of temperature on tree phenology and growth has gained particular attention in relation to climate change. While a number of reports indicate that warming can extend the length of the growing season and enhance tree growth rates, it is still debated whether temperature also affects biomass partitioning. Addressing the question of whether trees grown at different elevations invest similarly in various organs, we established four sites along an elevational gradient (320 to 595 m a.s.l.) in managed Norway spruce (Picea abies (L.) Karts) stands regenerating after clearcuts in central Norway. There, differences in temperature, bud break, tree growth, and allometric scaling were measured in small spruce trees (up to 3 m height). The results showed that bud break and shoot growth are affected by temperature, as lower sites completed the bud break process 5 days earlier than the higher sites did. There was some evidence indicating that the summer drought of 2018 affected tree growth during the season, and the implications of this are discussed. The allometric scaling coefficients did not change for the crown volume (slope value range 2.66–2.84), crown radius (0.77–0.89), and tree diameter (0.89–0.96) against tree height. A slight difference was found in the scaling coefficients of crown length against tree height (slope value range 1.04–1.12), but this did not affect the general scaling of the crown volume with tree height. Our results showed that different local environmental conditions affect both the growth rate and phenology in Norway spruce trees but, on the contrary, that the biomass partitioning among different parts of the tree remains essentially unchanged. This demonstrates that the allometric approach is an important tool for unraveling true vs. apparent plant plasticity, which in turn is an essential awareness for predicting plant responses to environmental changes.


Introduction
In temperate and boreal ecosystems, trees shift between growth and dormancy in cycles that are synchronized with seasonal environmental change, allowing survival in unfavourable seasons. These cycles are defined by physiological processes occurring in different plant tissues in response to local variations in temperature, day length, and light in terms of quantity and quality [1][2][3][4]. In addition, the magnitude and direction of phenological response to changes in temperature can be species-specific and/or site-specific [1,5]. Moreover, the uncertainty pertaining to climate change in terrestrial ecosystems calls for a better understanding of the thermal regulation of tree phenology and growth [5][6][7][8].
Forests 2020, 11, 1231 3 of 18 Furthermore, by studying the variability of allometric trajectories within juvenile stands, we aim to unravel the variability of allocation patterns in young trees, an under-represented aspect in the literature since most studies compare the allometric scaling of juvenile stands to mature forests [31,[48][49][50][51]. Therefore, we provide observational data that can improve our current understanding of environmental influence over a critical phase of stand development in a key European tree species.
In order to investigate possible variations in growth and allometric scaling trajectories due to temperature, we measured Norway spruce saplings along an elevational gradient in a northern continental climate. To quantify the differences in growing conditions, we observed the bud break process and tree growth throughout the season in relation to measured temperatures in selected plots. Further, we investigated whether the trees modified their structural balance in response to environmental variations by comparing allometric scaling trajectories.
Therefore, we aimed to test: • The influence of temperature on bud break and tree growth in order to quantify differences in the growing conditions along the gradient; • The stability of allometric trajectories with ontogenesis along the gradient.
We expect a positive effect of temperature on bud break and tree growth. The evidence behind this expectation comes from existing literature on the subject, which agrees on identifying temperature as the main environmental cue influencing bud break [1,2,12,13].
We also expect that allometric relationships, given their fundamental importance in determining tree fitness, are constant along the gradient. This hypothesis is based on the general quantitative theory of forest structure and dynamics [38,39] and is backed by existing studies providing evidence of the stability of allometric relationships despite environmental variations [33,[35][36][37].

Experimental Set
The study area (61 • 25 29.9" N, 11 • 04 46.6" E) is located on a hillslope in the Glomma river valley, Inland county, Norway. The climate is continental, with a precipitation peak during the summer and generally low temperatures, resulting in a boreal biome ( Figure 1). The conditions during data collection in the summer of 2018 were exceptionally warm and dry due to a severe drought that happened in Central and Northern Europe at the time. The event was quantified as "unprecedented" by Buras et al. [52]. In Norway, the season was described as "the hottest and driest summer since registrations started back in year 1900" by the Norwegian Meteorological Institute [53]. Available data suggest a strong influence of the drought on the study area, with the mean temperatures in May and July being more than 5 • C higher than average and the precipitation levels in July and August being more than 50 and 30 mm, respectively, lower than usual ( Figure 1).
The field study was set up in mono-layered stands, dominated by Norway spruce and managed with a production purpose. Study sites were distributed among stands that underwent clearcutting between 2010 and 2011 and were planted with saplings in 2012: site A (320 m a.s.l.), B (420 m a.s.l.), C (500 m a.s.l.), and D (595 m a.s.l.).
The varieties of Norway spruce planted in the sites were "Opsahl" in sites A, B, and C and "Kaupanger" in site D. Opsahl was selected to maximize stand yield in the climatic conditions of the area. Seeds belonging to this variety were collected from stands with an average latitude of 61.11 • N [54]. Kaupanger, on the other hand, was selected for its resistance to frost, particularly because of the anticipated dormancy in autumn. The average latitude of origin was 61.23 • N [55]. In the field, we were not able to distinguish between individuals of natural and artificial origins. We accounted for this possible source of variation by including tree variety in our models, which allowed us to test its effect on growth and phenology at the site level. Long-term average values for the period 1961-1990 ("long") are shown together with the mean values recorded in 2018 ("2018"). Temperature data were measured at Evenstad's weather station (4 km from study area), precipitation data were measured at Rena's weather station (~30 km south of study area). Source of the data: Norwegian Meteorological Institute (www.yr.no).
The field study was set up in mono-layered stands, dominated by Norway spruce and managed with a production purpose. Study sites were distributed among stands that underwent clearcutting between 2010 and 2011 and were planted with saplings in 2012: site A (320 m a.s.l.), B (420 m a.s.l.), C (500 m a.s.l.), and D (595 m a.s.l.).
The varieties of Norway spruce planted in the sites were "Opsahl" in sites A, B, and C and "Kaupanger" in site D. Opsahl was selected to maximize stand yield in the climatic conditions of the area. Seeds belonging to this variety were collected from stands with an average latitude of 61.11° N [54]. Kaupanger, on the other hand, was selected for its resistance to frost, particularly because of the anticipated dormancy in autumn. The average latitude of origin was 61.23° N [55]. In the field, we were not able to distinguish between individuals of natural and artificial origins. We accounted for this possible source of variation by including tree variety in our models, which allowed us to test its effect on growth and phenology at the site level.

Local Performance
To measure phenology and growth during the growing season, we established five plots for each site. The plots were established sparsely to account for possible local variations within the stand, while following these criteria: (i) areas with rocky or steeper soil were avoided, trying to limit the difference in soil water availability; (ii) each plot was more than 15 m from the border with the nearest forest stand. This was done to limit the edge effect of the nearby mature forest ecosystem due to shading and belowground competition.
After the determination of each plot center, we selected 10 juvenile trees among healthy individuals (showing no signs of desiccation or missing apical bud) in the vicinity of the center. Five plots were selected for each site, yielding 50 trees per site and a total of 200 trees across the sites. These individuals were marked with red tape and given IDs. This was done to perform repeated measurements of the following growth processes: Long-term average values for the period 1961-1990 ("long") are shown together with the mean values recorded in 2018 ("2018"). Temperature data were measured at Evenstad's weather station (4 km from study area), precipitation data were measured at Rena's weather station (~30 km south of study area). Source of the data: Norwegian Meteorological Institute (www.yr.no).

Local Performance
To measure phenology and growth during the growing season, we established five plots for each site. The plots were established sparsely to account for possible local variations within the stand, while following these criteria: (i) areas with rocky or steeper soil were avoided, trying to limit the difference in soil water availability; (ii) each plot was more than 15 m from the border with the nearest forest stand. This was done to limit the edge effect of the nearby mature forest ecosystem due to shading and belowground competition.
After the determination of each plot center, we selected 10 juvenile trees among healthy individuals (showing no signs of desiccation or missing apical bud) in the vicinity of the center. Five plots were selected for each site, yielding 50 trees per site and a total of 200 trees across the sites. These individuals were marked with red tape and given IDs. This was done to perform repeated measurements of the following growth processes: Our study design limited us to certain tree sizes and prevented us from performing a random sampling of trees, therefore sampled individuals should not be considered representative of the whole site in terms of tree size.
The values of apical shoot growth and diameter increment were normalized over tree height and diameter, respectively, as recorded at the start of the season. This allowed us to evaluate and compare the growth rate of trees of different dimensions during the season. The height and diameter values used for the normalization were measured at the start of the growing season.
Apical stages of bud break were classified using a phenological scale dedicated to spruce, adapted from Fløistad and Granhus [56]: (0) dormant buds; (1) buds slightly swollen; (2) buds swollen, bud  [56], ours is shorter and relies only on qualitative aspects. We removed stages relying exclusively on quantitative criteria, as they are inaccurate in evaluating bud break under field conditions. At each plot, temperature loggers were installed, yielding a total of 20 loggers. We used HOBO 8K pendant ® waterproof temperature loggers (Onset Computer Corporation, 470 MacArthur Blvd., Bourne, MA, USA), each placed inside a white solar shield case specifically built to shelter the logger from the influence of direct sunlight while also granting aeration. Temperature data were logged every 10 min. Due to logistic problems, the loggers were installed in the field as late as 3 July, well into the growing season. We recorded data from the 3 July until 24 September. We obtained the missing data (15 May to 3 July) by imputation, modelling the relationship between each logger and the Evenstad weather station, which was located 4 km away from the sites. The specific coefficients defining these relationships are available in Table S1 of the Supplementary Materials.

Allometric Relationships
The collection of allometric data occurred in the same sites (i.e., the same clearcut stands) as the measurement of bud break and growth.
We sampled trees varying from 10 to 300 cm in height and divided them into six groups on the basis of height: 0-50, 50-100, 100-150, 150-200, 200-250, and 250-300 cm. We sampled 10 trees per height interval, resulting in 60 trees per site and a total of 240 trees. Each tree was only sampled once; we sampled all sites between 19 and 24 September.
Because of this height range constraint, the trees sampled for allometric data included, but were not limited to, individuals on which we performed phenological and growth analyses. Plants were sampled based on their height and health status (no visible sign of desiccation, no broken branches or missing tips). We took the following measurements: (1) tree height; (2) tree base diameter; (3) crown radius (R cro )-i.e., the distance between the tip of the longest branch and the stem; (4) crown insertion height-i.e., the distance between the lowest active branch in the canopy and the ground. From these, we obtained also (5) crown length (L cro )-i.e., the distance between the lowest active branch and the apical tip-and (6) crown volume (V cro ), obtained by multiplying the crown length by the square of the crown radius (V cro = L cro × R cro 2 ). As in previous studies [35,36], this simple formula is preferred over other approaches to estimate crown volume because it easily allows one to investigate the relative change in crown volume-namely, the scaling exponent of V cro -with tree height. The use of more complicated formulas to approach crown shapes (e.g., cone) does not improve the estimation of scaling parameters.

Statistical Analyses
All the statistical analyses were performed using the R software, version 3.6.1 [57]. We first tested for significant differences in the growing conditions between different sites. We applied analysis of variance (ANOVA) to test for differences in temperature, phenology, and growth. For temperature, we compared the mean of hourly temperatures. For phenology, we considered the day (in progressive "day of the year" DOY values) on which the final stage of bud break (stage 5) was reached. For growth, we considered the total apical shoot elongation and total diameter increment values, normalized over tree height and tree diameter, respectively.
We applied either the parametric ANOVA or Kruskal-Wallis tests depending on the data distribution. Normality in the data and residual distribution was tested with the Shapiro-Wilk test.
The homogeneity of variance was tested with the Bartlett test.
Since the ANOVA only reveals that "at least" one site is different from the others, in the case of significant (p < 0.05) response we applied the Tukey (for parametric ANOVA) or Dunn (for Kruskal-Wallis) post hoc tests to determine specific differences between sites (complete statistical outputs available in Table S2 of the Supplementary Materials). Additionally, we fitted mixed-effect models on phenology and growth to test the influence of temperature, daylength, date, and variety over the processes of bud break (including all observed stages) and apical shoot elongation (including all elongation measurements). In that respect, the temperature values from each plot were linked to the response variables measured from the saplings in the plots. For the mixed-effect models, we included plot (i.e., each plot's ID) as a random term when it improved the model and thus accounted for non-measured local environmental variations. We applied cumulative link mixed-effect models (R package "ordinal" [58]) to phenological-scale values, testing for the effect of temperature, daylength, and variety. We applied linear mixed-effect models (R packages "lme4" [59] and "lmerTest" [60]) to apical growth values, testing for the effect of temperature, date, and variety. We tested whether the inclusion of a random variable improved the models by comparing the AIC (Akaike Information Criterion) values of the mixed-effect models with that of models which did not include random variables (glm for continuous variables and clm for ordinal variables). Finally, we performed an AIC-based model selection using the dredge automated model selection function in the "MuMIn" package [61].
Repeated measurements of tree diameter resulted in being very prone to measurement errors because of the field conditions and small size scale of our trees. In order to reduce the error, we only used the first and last measurements to obtain the total diameter increment for the whole season. This allowed us to test for differences between the sites using ANOVA, but the full data series was not reliable for mixed-effect modelling.
Finally, we investigated allometric scaling to assess whether differences in growth are reflected in differences in the structural balance of the trees. We considered the scaling of crown volume, crown radius, crown length, and tree base diameter over tree height. All the scaling parameters were transformed by a base 10 logarithmic function, which allowed us to model the allometric relationship as a linear regression (Equation (1)) [62].
where y and x are the structural traits being considered (e.g., crown volume and tree height) and b is the scaling exponent and a is the intercept-i.e., the coefficients that define the relationship. This approach allowed us to the quantify site-specific values for the allometric scaling exponent (b). Comparing b values, we were able to investigate the variation in allometric trajectories along the elevational gradient-i.e., the structural balance between traits-in different sites. Similarity or differences between sites are expressed by the confidence intervals (C.I.) of the b parameter; overlapping C.I. values for the b parameters indicate similarity.

Local Performance
The ANOVA revealed significant (p < 0.001, χ 2 =21.158, df = 3) temperature differences between the sites. The post hoc Dunn test identified the two lower (A and B) and the two higher (C and D) sites as statistically different (Tables 1 and 2), with higher mean temperatures at lower elevations. More specifically, site A was warmer than sites C (p < 0.001) and D (p = 0.002). Similarly, site B was warmer than sites C (p = 0.005) and D (p = 0.018). Table 1. Results of the post hoc test used to assess the differences in hourly temperature ( • C) between sites for the growing season of 2018 (15 May to 4 September). "Group" letters identify significantly (p < 0.05) different groups, where "a" is the group with the lower value. Corresponding mean temperature ( • C) values for the growing season are also shown for each site. Values in brackets represent the standard error (S.E.).

Site
Group  .
The tree size variables varied in a similar fashion, as the ANOVA results identify significant differences between sites also for height (p < 0.001, F = 13.4, df = 3) and diameter (p < 0.001, χ 2 = 26.232, df = 3) values, measured at the start of the season (23 May). Post hoc tests revealed differences between the two lower and two higher sites, with trees being smaller at higher elevations ( Table 3). The trees in sites A and B resulted in being higher than the trees in sites C and D, with p-values < 0.001. The trees in site A also had larger diameters those in than sites C (p < 0.001) and D (p = 0.009). Similarly, individuals in site B had larger diameters than those in sites C (p < 0.001) and D (p = 0.004). Table 3. Results of the post hoc tests used to assess the differences in tree height and diameter in all four sites. Values were measured at the start of the growing season (23 May). "Group" letters identify significantly (p < 0.05) different groups, where "a" is the group with the lower value. Corresponding mean height (cm) and diameter (cm) values are also shown for each site. Values in brackets represent the standard error (S.E.). To compare the timing of growth onset across sites, we considered the day of completion of the bud break process-namely, the date in which each individual reached stage 5 of the phenological scale. It took 13 days, from 18 May to 31 May, for all individuals in sites A and B to complete the bud break process. This happened 5 days earlier than in sites C and D, where the process was completed on 5 June (Figure 2). To compare the timing of growth onset across sites, we considered the day of completion of the bud break process-namely, the date in which each individual reached stage 5 of the phenological scale. It took 13 days, from 18 May to 31 May, for all individuals in sites A and B to complete the bud break process. This happened 5 days earlier than in sites C and D, where the process was completed on 5 June (Figure 2). The mean date of completion of the bud break process is only one to two days earlier in lower sites A and B (Table 4). Nevertheless, applying ANOVA revealed significant (p = 0.0015, χ 2 = 15.395, df = 3) differences between the sites. The post hoc analysis results indicate an earlier mean date of bud break in site A than in site D (p = 0.017) and in site B than sites C (p = 0.021) and D (p = 0.003) ( Table  4). Table 4. Results of the post hoc test used to assess differences in the bud break process in the study sites. The parameter being tested is the mean date of completion of the bud break process (in day of the year values). "Group" letters identify statistically different groups, where "a" is the group with the lower value. Dates are expressed as "day of the year" (DOY) progressive values, the corresponding dd/mm dates are also shown. Values in brackets represent the standard error (S.E.).

Site
Group  The mean date of completion of the bud break process is only one to two days earlier in lower sites A and B (Table 4). Nevertheless, applying ANOVA revealed significant (p = 0.0015, χ 2 = 15.395, df = 3) differences between the sites. The post hoc analysis results indicate an earlier mean date of bud break in site A than in site D (p = 0.017) and in site B than sites C (p = 0.021) and D (p = 0.003) ( Table 4). Table 4. Results of the post hoc test used to assess differences in the bud break process in the study sites. The parameter being tested is the mean date of completion of the bud break process (in day of the year values). "Group" letters identify statistically different groups, where "a" is the group with the lower value. Dates are expressed as "day of the year" (DOY) progressive values, the corresponding dd/mm dates are also shown. Values in brackets represent the standard error (S.E.). The CLMM that was applied on the apical stage data included the random variable "plot", and indicated a significant (p < 0.001, Z = 28.15, S.E. = 0.03002) positive (0.845) correlation between temperature and bud break. Tree variety and daylength were discarded during model selection, indicating no influence of these variables over the bud break process.

Site
Tree growth differed significantly between sites according to ANOVA, in terms of both normalized apical shoot elongation (p = 0.0318, F = 3, df = 3) and normalized diameter increment (p = 0.0014, χ 2 = 15.51, df = 3). The differences in apical growth are limited, as the only significant difference is between sites A and C (p = 0.048), with site A growing comparatively less than site C. The differences in diameter increment are stronger, as the normalized radial growth in site A was lower than in sites C (p = 0.0138) and D (p = 0.0026), and lower in site B than in site D (p = 0.03) ( Table 5). These differences do not follow the same elevational pattern observed for temperature, tree size, and phenology; when looking at normalized growth values, trees in lower sites grew similarly or even less than those at higher elevations. Table 5. Results of the post hoc tests used to assess the differences in tree growth during the season. Parameters being tested were the mean normalized values of apical shoot elongation and diameter increment. Mean values, both normalized and absolute (cm), are also shown for each site. "Group" letters identify statistically different groups, where "a" is the group with the lower value. Values in brackets represent the standard error (S.E.). The LME (linear mixed effect) model applied to apical shoot growth for the whole growing season showed a significant positive influence of temperature on growth (coeff = 14.1238, p = 0.00904, t = 2.619, S.E. = 5.39236). The inclusion of "date" improved the model but did not have a significant effect (p = 0.16104, t = 1.403, S.E. = 0.52593). Date, however, had a significant negative interaction with temperature (coeff = −0.07035, p = 0.01252, t = −2.505, S.E. = 0.02809), indicating that the influence of temperature on apical shoot growth decreased over time. Including "plot" as a random variable accounting for environmental variation improved the model, while the "variety" variable was discarded during model selection.

Site
Unexpectedly, the diameter increment during the season displayed a peaks-and-trough pattern. This pattern was found in all four sites, but the trough was deeper in lower sites (Figure 3). In the three-week period from 13 June to 4 July, the average normalized diameter increment values were lower in all sites if compared with the previous and following three-week periods. This means that the growth either slowed down or stopped altogether in the middle of the season. The normalized apical shoot growth values do not show any interruption and simply decrease to a stop, with most of the elongation being over by 4 July (Figure 3).

Allometric Relationships
The slope coefficients in the linear regression function, modelling the allometric scaling of the crown volume vs. tree height, did not show significant differences among sites (Figure 4).

Allometric Relationships
The slope coefficients in the linear regression function, modelling the allometric scaling of the crown volume vs. tree height, did not show significant differences among sites (Figure 4).   Further modelling of the allometric scaling of structural parameters other than the crown volume supports this result, with the allometric trajectories being consistently similar in all sites for the values of tree diameter (D), crown length (L cro ), and crown radius (R cro ) over tree height (H) ( Table 6). One single exception is the L cro vs. H relationship in site A (Table 6), whose b value is significantly higher than in other sites.

Discussion
Our results are in line with the present literature concerning the role of temperature in regulating bud break in northern tree species [1,2] and provide detailed modelling of this process under field conditions. This increases the strength of the current projections, associating an earlier growth onset with the increasing temperatures linked to climate change [5][6][7].
Similarly, we expected to see higher growth rates in warmer sites, but the normalized values of apical and radial growth appear irregular and do not follow the same pattern as the elevational temperature gradient. When looking at the radial increment values (Figure 3), our data point to a slowing in the tree growth rate during the season, which contrasts with existing knowledge [63] and possibly indicates the influence of the year's drought.
Although we lack the necessary data to determine the drivers behind these patterns, we argue that the observed significant differences in growth rates point to differences in the growing conditions between sites. Despite these variations, we find that the exponents of allometric trajectories are similar in the four sites. This provides evidence that spruce saplings maintain the same trajectories of structural balance during growth, regardless of the local growing conditions. The measured temperatures along the elevational gradient indicate two distinct local climates: a "lower" area, comprising sites A and B and ranging from 320 to 420 m a.s.l., and a "higher" area comprising sites C and D and ranging from 500 to 595 m a.s.l. (Tables 1 and 2). The sampled trees in these sites seem to mirror these conditions, with trees in low-elevation sites (A-B) being significantly bigger than those at higher elevations (C-D) both in height and diameter (Table 3). Similarly, the positive influence of temperature on bud break caused saplings at the lower, warmer sites A and B to experience a faster growth onset. Lower sites showed a faster rate of completion (perc. of trees reaching the final stage) during the whole period (Figure 2), and completed the process 5 days earlier than the higher sites did. Differences in the mean day of completion of the bud break process are limited, amounting only to one to two days, but still indicate a later start of the season at higher elevations (Table 4). This is consistent with existing knowledge, as temperature is the main environmental cue influencing bud break in northern tree species [1,2] and temperatures decrease with increasing elevation, resulting in delayed bud break [9][10][11].
Despite the variation in phenology across elevation, the measured values of tree growth did not follow such a pattern, as the growth rate variables were generally of smaller value at lower elevation, and the differences between the sites were much less pronounced. The values of normalized apical shoot elongation appear to be significantly different only between sites A and C, with trees in the lowest site A producing comparatively shorter shoots. Differences in the normalized diameter increment are slightly more evident (Table 5), but again indicate that trees in lower sites had a comparatively lower radial increment than trees in higher sites. Overall, it appears that bigger trees in lower sites (A and B) grew comparatively less than smaller trees in higher sites (C and D) during the season. This appears to be in contrast with the well-known direct correlation between temperature and growth rate, as confirmed by our modelling. This inconsistency could be explained, at least for radial growth values, by the pattern of measured values during the season (Figure 3). Rossi et al. [63] found that conifers in cold climates increase their radial growth rate with day length, with a peak at the day of maximum day length. This would be the summer solstice, 21 June 2018-i.e., at the time of the season when we recorded the lowest radial growth rates (Figure 3). This discrepancy is likely due to the summer drought of 2018 [52,53]. Severe effects of the drought on forests were observed in Central Europe during the same event [64], and similar or worse effects likely happened in Scandinavia, where deviations from the norm where stronger [52].
The warm and dry conditions of the summer may have caused the peak-and-trough pattern either by slowing down the tree growth or by causing the shrinking of the xylem cells due to dehydration [65]. It appears that lower-elevation sites were more affected by this event (Figure 3). This would explain the comparatively higher radial growth values observed in the higher-elevation sites C and D (Table 5), where lower temperatures may have diminished the drought effect. This is an untested hypothesis, since assessing and quantifying the local impacts of the drought was not within the scopes of our study and would need further investigation. Furthermore, the fact that our measurements cover such a small area makes it difficult to generalize our results. However, given the existing knowledge on the effects of the 2018 summer drought [52,64] and the prospected increase in the frequency of extreme events due to ongoing clime change [66], we consider this aspect of our study relevant enough to be discussed.
The growth rate pattern for the normalized apical shoot elongation values is consistently similar in all sites; after an initial peak, it slows down and almost stops by 4 July (Figure 3). The significant negative effect of the interaction between the temperature and date variables shows a decrease in the positive effect of temperature on growth over time. This is due to apical growth cessation during the season, while temperatures are still high. Unlike radial growth, there is no recovery in the growth rate after it slows down. This may be caused by the fixed growth pattern of P. abies; at the start of the season, the bud already contains all the cell primordia that will develop during the season [2]. As soon as the pre-determined number of cells has developed, apical shoot growth stops and the apical bud enters dormancy. We have no way to establish whether the apical shoot growth ceased because all of the bud primordia properly developed or because the drought forced an early dormancy of the apical buds.
Variety does not seem to have a significant effect on bud break or apical shoot elongation. This is unexpected, as the characteristics of the two varieties are very different. Trees planted in sites A, B, and C belong to the "Opsahl" variety and are selected to maximize stand yield [54,55]. Therefore, we expected to see some differences-e.g., higher growth rates in these sites. Our sampling involved both natural and artificial regeneration, so it is possible that we sampled few individuals of non-local origin and the influence of genotypes was not represented in the dataset. Another possibility is that the drought offset the effect of different varieties. If lower-elevation sites really grew less because of the drought, this may have counterbalanced the effect of the "Opsahl" variety. Again, it should be noted that this hypothesis is untested, but this should be discussed and investigated by future studies. The possible management implications for Scandinavian forestry in the face of climate change are huge, given the importance of Norway spruce in the area [19]. As the tree species distribution is projected to change dramatically [14,15], spruce forests will experience more frequent primary and secondary disturbances [22,24]. Observed and predicted increases in forest growth with global warming [67,68] could be disrupted by these events, as noted by other studies [69,70]. Our results seem to point in this direction, and highlight the need for more field-based studies to adapt future silviculture to the changing climate.
Although the variations in tree growth rates during the season do not reflect the elevational temperature gradient, they were significant nonetheless. This points to local differences in growing conditions, with factors other than temperature influencing growth onset and growth rates during the same season. Despite these differences, the scaling exponents that define the allometric relationships of structural parts of the trees do not vary significantly (Figure 4, Table 6). This means that crown volume, stem diameter, crown length, and crown radius grow similarly, in relation to tree height, in all four sites. The only exception appears to be the scaling relationship L cro vs. H (Table 6), as the scaling value C.I.s of site A do not overlap with other sites, being slightly higher. A higher scaling value of L cro vs. H means that the trees in site A tend to favor vertical growth over time. This is offset by the relatively lower scaling rate in site A for R cro vs. H ( Table 6). As a result, the scaling of V cro vs. H is the same in all sites (Figure 4). Therefore, it would appear that trees in site A develop comparatively narrower crowns over time.
The difference between the slope values of L cro vs. H is minimal, yet this is a noteworthy discrepancy since the common knowledge is that Norway spruce tends to grow narrower crowns in colder, harsher environments typically associated with higher elevations and latitudes [71,72]. The fact that this applies to the lowest and warmest site A is surprising. It was observed in the field that competition with Betula spp. is stronger at lower elevations, likely because of the higher temperatures. It is possible that this drives spruce saplings in site A to favor vertical growth in early development phases. Despite this, the general balance between productive organs (in our case, crown volume) and tree dimension (tree height) follows the same trajectory in all sites. An important aspect of our study is that the regeneration of both natural and artificial origin comes from a restricted range of latitudes (61.11 • N to 61.25 • N). Another study conducted by Anfodillo et al. [35] found that the scaling exponent of crown volume versus tree height does not vary significantly (2.22 to 2.31) between four temperate mountain forests with different composition and disturbance levels but within a restricted range of latitudes (46.06 • N to 47.27 • N). Sellan et al. [36] obtained a similar outcome when comparing the scaling exponents (3.30 to 3.38) of three tropical forest plots with different species richness and composition between 0 • and 1 • N. Therefore, latitude appears to be one major factor influencing the slope of the locally optimal allometric relationships between these traits.
The continuity in the measured scaling exponents supports the hypothesis that allometric relationships are fundamental to tree functioning and, therefore, that trees acclimate to local conditions by changing their growth rate while maintaining an optimal structural balance during onthogenesis [33]. From this perspective, allometric relationships are a major constraint that governs tree growth, and can be a powerful tool in modelling tree and forest dynamics. This is consistent with the general quantitative theory of allometry proposed by West, Brown, and Enquist [38,39]. It is fundamental to keep in mind that this is observed for the scaling of structural parts versus tree height. The scaling of traits versus diameter has shown more variability in response to environmental variation [30,40,73].

Conclusions
In this study, we quantified the influence of temperature on bud break and growth under field conditions. Temperatures follow an elevational gradient, being higher in lower sites and positively influencing the bud break process. Consequently, the season started earlier at lower elevation sites. Differences in growth appear less consistent with temperature, as the bigger trees in the lowest, warmest site A grew comparatively less than in the higher sites C and D. The values of diameter increment during the season point to a possible negative influence of the exceptional 2018 summer drought, which would explain this discrepancy.
The measured differences in temperatures and growth point to significant differences in local growing conditions. Despite this, the allometric relationships of crown volume and tree diameter versus tree height do not vary along the elevational gradient. This indicates that young spruce trees do not appear to modify their structural balance and biomass partitioning to acclimate to local conditions. The allometries of crown length and crown radius versus tree height are stable as well, with the exception of a minor difference in crown length in site A, which did not have an impact on the overall crown volume-height balance.
Our data support the hypothesis that maintaining specific allometric trajectories is fundamental for tree functioning, even at such a young age of tree and stand development. By comparing the growing conditions and allometric scaling in Norway spruce saplings, we offer insight into the ecological factors regulating its growth and allocation patterns. We show that the simple application of an allometric approach can provide valuable information on true vs. apparent plant plasticity, thereby increasing our understanding of tree and forest dynamics and our predictive ability in the face of climate change.
We recommend that future studies wishing to understand variations in tree biomass partitioning adopt a similar allometric-based perspective. On a broader scale, further investigations are needed to fully understand the drivers of allometric scaling in trees. Working along gradients is one efficient way of testing for the variations and drivers of tree and forest growth in the ongoing effort to formulate a comprehensive theory. Additionally, our results highlight the importance of field-based studies that allow for unforeseen changes in environmental conditions. As climate change intensifies, extreme events such as droughts are projected to increase in a diffused manner [66], and field studies can provide useful data on tree responses to such disturbances.
Supplementary Materials: The following are available online at http://www.mdpi.com/1999-4907/11/11/1231/s1: Table S1: Intercept and slope values defining the relationship between the temperature in each plot and the Evenstad weather station. These values were used to compute missing temperature values for the period ranging from 15 May to 3 July, Table S2: Results for post hoc tests performed to identify statistical differences between sites. "Comparison" columns indicate the two sites being compared. Significance levels for p-values: "***" 0.001, "**" 0.01, "*" 0.05. Z = Z value for Dunn test; p.unadj = unadjusted p-value for Dunn test; p.adj = p-value adjusted to multiple comparisons; diff = difference in average between compared sites; lwr = lower value for 95% C.I.; upr = upper value for 95% C.I.