Initial Survival and Development of Planted European Beech (Fagus sylvatica L.) and Small-Leaved Lime (Tilia cordata Mill.) Seedlings Competing with Black Cherry (Prunus serotina Ehrh.)

Black cherry (Prunus serotina Ehrh.) is considered one of the most invasive tree species in central Europe and causes problems for both nature conservation and silviculture. Besides mechanical control treatments, a suggested control method to prevent its ongoing spread is to underplant shade-tolerant native tree species. Therefore, we combined two mechanical treatments, with underplanting of European beech (Fagus sylvatica L.) or small-leaved lime (Tilia cordata Mill.) on fenced and unfenced plots. After the first growing season, survival rates were evaluated, and selected seedlings were destructively harvested to analyze their growth performance and leaf morphology in association with the different light regimes resulting from mechanical treatments Survival rates for both seedlings were very high (>95%). Survival rates were higher on fenced plots than on unfenced plots, most likely as result of browsing. The mortality of F. sylvatica decreased with increasing light availability on fenced plots. The mortality of T. cordata did not change along the light gradient. After one vegetation period no differences with respect to biomass allocation could be detected along the light gradient. However, the specific leaf areas of both species responded similarly, decreasing with increasing light availability. In summary, both species were able to establish and survive in the dense P. serotina understory and might have the potential to outcompete the invasive alien species in the long run.


Introduction
Biotic invaders can inflict enormous environmental damage to ecosystems by altering the integrity of their fundamental properties [1]. Invasive plant species affect competition for resources like water or nutrients, and can change the invaded ecosystems' components such as increased biomass production or nitrogen availability and litter production with higher decomposition rates for instance [2]. They are globally recognized as major threats to biodiversity, nature conservation, and ecosystem stability in native forests [1,3]. The urgent question, when dealing with any kind of invasive species, is how to handle or control them after their presence is discovered to avoid or minimize possible damage to the ecosystem.
The deciduous tree species black cherry (Prunus serotina Ehrh.), native to North America [4], is the largest cherry in its native range [5] and is a valuable furniture wood in the United States [6]. P. serotina biomass allocation of underplanted tree-seedlings under different light conditions in order to evaluate their growth performance.
The aims of this survey were (a) to study whether the chosen underplanted tree species survive in the dense P. serotina understory, (b) to relate biomass allocation patterns in the small seedlings to light availability, and (c) to relate specific leaf area to the light available as a result of previously conducted mechanical control-methods applied to the invasive species.

Seedling Mortality Inside and Outside the Fences
After the first growing season 2017, the overall mortality of both species was negligible (Table S1 in supplementary material contains a complete dataset on dead seedlings). Of the 2108 F. sylvatica seedlings, 92 died and of the 2062 T. cordata seedlings, 94 died, which equaled an overall mortality of 4.4% and 4.6% and indicated no differences in mortality between species (Chi-squared; p = 0.819). For both species, the number of dead seedlings was higher on unfenced plots compared to the fenced plots (F. sylvatica: p = 0.038 and T. cordata: p = 0.005). No difference between the two species could be detected within (p = 0.946) or outside the fences (p = 0.643) ( Figure 1). concomitantly their access to limiting resources such as nutrients, water, and light [29]. The present study focused on biomass allocation of underplanted tree-seedlings under different light conditions in order to evaluate their growth performance. The aims of this survey were (a) to study whether the chosen underplanted tree species survive in the dense P. serotina understory, (b) to relate biomass allocation patterns in the small seedlings to light availability, and (c) to relate specific leaf area to the light available as a result of previously conducted mechanical control-methods applied to the invasive species.

Seedling Mortality Inside and Outside the Fences
After the first growing season 2017, the overall mortality of both species was negligible (Table S1 in supplementary material contains a complete dataset on dead seedlings). Of the 2108 F. sylvatica seedlings, 92 died and of the 2062 T. cordata seedlings, 94 died, which equaled an overall mortality of 4.4% and 4.6% and indicated no differences in mortality between species (Chi-squared; p = 0.819). For both species, the number of dead seedlings was higher on unfenced plots compared to the fenced plots (F. sylvatica: p = 0.038 and T. cordata: p = 0.005). No difference between the two species could be detected within (p = 0.946) or outside the fences (p = 0.643) ( Figure 1).

Mortality along light gradient
On average, one seedling died on each plot over the entire field experiment. Note that a single dead seedling represented a mortality of 3.13% (Table 1). The fenced and unfenced plots did not differ in light availability (p = 0.294). Thus, the further analysis of mortality and growth performance (seedlings' biomass, compartment mass fraction analysis, and specific leaf area) was restricted to the fenced sites, to exclude an effect of browsing. The correlation between mortality and light availability was only significant for F. sylvatica seedlings, with mortality decreasing with increasing light availability (Figure 2a). No correlation between mortality and the light availability (ISF, indirect site factor, Table S3 in supplementary material contains all abbreviations and definitions used in this manuscript) was detected for T. cordata seedlings and the calculated GAM (generalized additive model) was also not significant ( Figure 2b).

Mortality along light gradient
On average, one seedling died on each plot over the entire field experiment. Note that a single dead seedling represented a mortality of 3.13% (Table 1). The fenced and unfenced plots did not differ in light availability (p = 0.294). Thus, the further analysis of mortality and growth performance (seedlings' biomass, compartment mass fraction analysis, and specific leaf area) was restricted to the fenced sites, to exclude an effect of browsing. The correlation between mortality and light availability was only significant for F. sylvatica seedlings, with mortality decreasing with increasing light availability (Figure 2a). No correlation between mortality and the light availability (ISF, indirect site factor, Table S3 in supplementary material contains all abbreviations and definitions used in this manuscript) was detected for T. cordata seedlings and the calculated GAM (generalized additive model) was also not significant ( Figure 2b). Table 1. Mortality per plot for F. sylvatica and T. cordata seedlings after the first growing season. Shown are detected minimum (Min), mean ± standard deviation (SD), median, and maximum (Max) mortality of dead seedlings on fenced and unfenced plots after the growing season 2017 and associated figures (Ass. figure).

Species
Fencing Plots (n) Mortality Per Plot (%) Ass. Figure  Min Mean ± SD (Median) Max    (a) (b) Figure 2. Mortality of (a) F. sylvatica seedlings (FS) and (b) T. cordata seedlings (TC) on fenced plots (n = 33) in association with light, measured as indirect site factor (ISF). Significant trends (p < 0.05) in association with light are shown as a black line, non-significant trends as a grey line. Shaded grey area shows the 95% confidence interval. Both explained deviance and p-values were extracted from the generalized additive model (GAM)-models.

Biomass before and after Growing Season
Due to browsing damage on a side shoot, one F. sylvatica seedling had to be excluded in this analysis. The seedlings' dry mass, calculated as total dry mass (TDM), aboveground biomass (AGB), and root mass (RM), were compared before and after the first growing season for both tree species. Changes could not be detected, neither in F. sylvatica (Kruskal-Wallis-Test: 0.07 < p < 0.11) nor in T. cordata (0.17 < p < 0.24) ( Table 2).

Figure 2.
Mortality of (a) F. sylvatica seedlings (FS) and (b) T. cordata seedlings (TC) on fenced plots (n = 33) in association with light, measured as indirect site factor (ISF). Significant trends (p < 0.05) in association with light are shown as a black line, non-significant trends as a grey line. Shaded grey area shows the 95% confidence interval. Both explained deviance and p-values were extracted from the generalized additive model (GAM)-models.

Biomass before and after Growing Season
Due to browsing damage on a side shoot, one F. sylvatica seedling had to be excluded in this analysis. The seedlings' dry mass, calculated as total dry mass (TDM), aboveground biomass (AGB), and root mass (RM), were compared before and after the first growing season for both tree species. Changes could not be detected, neither in F. sylvatica (Kruskal-Wallis-Test: 0.07 < p < 0.11) nor in T. cordata (0.17 < p < 0.24) ( Table 2). Table 2. Seedlings' biomass before and after the growing season 2017. Total dry mass (TDM), aboveground biomass (AGB), and belowground biomass (RM) of F. sylvatica and T. cordata before (nursery) and after the first vegetation period (harvest) in g. The leaf weights from harvested seedlings were excluded from TDM and AGB. n 1 is the sample size of seedlings from the nursery, n 2 is the sample size of the harvested plants. Shown are minimum (Min), mean and standard deviation (SD), median, and maximum (Max) values. The three tested stands (b2, c2, c3) did not reveal a significant difference on the mass fractions (MF) of leaves (LMF), branches (BMF), stem (SMF), and roots (RMF) neither for F. sylvatica (0.14 < p < 0.70) nor T. cordata (p = 0.06 < p < 0.65). The treatment of P. serotina itself also showed no significant difference on the compartment mass fractions for F. sylvatica (0.82 < p < 0.94) or T. cordata (0.26 < p < 0.55). Table S2 (supplementary material) contains the complete seedlings dataset on mass fractions.

Species
The LMF of F. sylvatica seedlings was the only mass fraction that showed a significant trend along the light gradient ( Table 3). The LMF increased by 0.04% with 1% light availability, measured as ISF ( Figure 3). The three other mass fractions in F. sylvatica (BMF, SMF, RMF) and also AGB did not respond along the gradient (0.127 < p < 0.310) ( Table 3). AGB and none of the mass fractions of T. cordata showed a significant trend along the light gradient (0.127 < p < 0.841) ( Table 3). Table 3. Results from generalized linear models (GLM) for seedlings growth response analysis in regard to light (model: y~ISF). Results for F. sylvatica and T. cordata with light (ISF) as influencing factor on aboveground biomass (AGB), leaf mass fraction (LMF), branch mass fraction (BMF), stem mass fraction (SMF), root mass fraction (RMF), and specific leaf area (SLA).

Species
Response
The LMF of F. sylvatica seedlings was the only mass fraction that showed a significant trend along the light gradient ( Table 3). The LMF increased by 0.04% with 1% light availability, measured as ISF ( Figure 3). The three other mass fractions in F. sylvatica (BMF, SMF, RMF) and also AGB did not respond along the gradient (0.127 < p < 0.310) ( Table 3). AGB and none of the mass fractions of T. cordata showed a significant trend along the light gradient (0.127 < p < 0.841) ( Table 3).  Table 3. Results from generalized linear models (GLM) for seedlings growth response analysis in regard to light (model: y ~ ISF). Results for F. sylvatica and T. cordata with light (ISF) as influencing factor on aboveground biomass (AGB), leaf mass fraction (LMF), branch mass fraction (BMF), stem mass fraction (SMF), root mass fraction (RMF), and specific leaf area (SLA).

Species
Response Figure   Figure 3. Leaf mass fraction (LMF) of F. sylvatica seedlings (FS) from fenced plots in association with indirect site factor (ISF). The sample-size was n = 98. Black solid line shows the trend and the shaded grey area shows the 95% confidence interval. R 2 and the p-value were extracted from the GLM (generalized linear model).
SLA was negatively correlated with light in both tree species, as SLA decreased with an increase in light (both species: p < 0.001) ( Table 3). A greater proportion of the variance in SLA was explained by light in leaves of T. cordata (R 2 = 0.21) in comparison to leaves of F. sylvatica (R 2 = 0.13) ( Table 3).
A comparison of the slopes in both models revealed significant differences (p = 0.02) for the two species along the gradient. The response of leaves to increased light was stronger for leaves of T. cordata, indicating higher morphological plasticity than for F. sylvatica (Figure 4). Sample size for F. sylvatica was n = 96 and for T. cordata n = 99. The trends were significant for both tree species (p < 0.001) and are shown as solid lines (black shows the trend for T. cordata and grey shows the trend for F. sylvatica). Dashed lines show the 95% confidence intervals.

Higher Mortality outside the Fence Due to Herbivory
Mortality after the first vegetation season was marginal ( Figure 1). Therefore, the seedlings' initial survival was considered a successful start of the experiment. The low losses were potentially also due to the hydrogel treatment applied before planting which may have functioned as a

Specific Leaf Area
Three harvested F. sylvatica seedlings with only dried-out leaves had to be excluded in this analysis. The mean leaf area (LA) did not respond along the light gradient (F. sylvatica p = 0.936; T. cordata p = 0.658). The specific leaf area (SLA) in F. sylvatica leaves ranged from 170.7 to 307.0 cm 2 g −1 , which was on average less than for T. cordata leaves ranging from 190.6 to 444.8 cm 2 g −1 . SLA was on average 25.3% greater in leaves of T. cordata (p < 0.001) ( Table 4).  SLA was negatively correlated with light in both tree species, as SLA decreased with an increase in light (both species: p < 0.001) ( Table 3). A greater proportion of the variance in SLA was explained by light in leaves of T. cordata (R 2 = 0.21) in comparison to leaves of F. sylvatica (R 2 = 0.13) ( Table 3).
A comparison of the slopes in both models revealed significant differences (p = 0.02) for the two species along the gradient. The response of leaves to increased light was stronger for leaves of T. cordata, indicating higher morphological plasticity than for F. sylvatica (Figure 4).

Higher Mortality outside the Fence Due to Herbivory
Mortality after the first vegetation season was marginal ( Figure 1). Therefore, the seedlings' initial survival was considered a successful start of the experiment. The low losses were potentially also due to the hydrogel treatment applied before planting which may have functioned as a mechanical protection for the fine roots [31]. The finding that more seedlings died outside the fence (Figure 1) can be explained by grazing mammals, whose browsing on seedlings commonly result in the loss of the leader shoot [32]. This loss is closely correlated to seedling mortality if approximately one third of the shoot length is browsed [33]. In the present study seedlings were sometimes browsed down to ground level or were partly missing. During the time this field experiment, grazing species such as roe deer (Capreolus capreolus, L. 1758), red deer (Cervus elaphus, L. 1758) [34], wild boar (Sus scrofa, L. 1758) [35], and European hares (Lepus europaeus, PALLAS 1778) [36] were recorded by cameras directly on the study sites. Browse damage is most intense on seedlings of 1.00-1.30 m in height [37]. The underplanted seedlings in the stands in Linde (min. size 0.16 to max. size 1.24 m) were therefore especially susceptible to grazing by mammalian herbivores.

F. sylvatica Mortality Response to Light
Shade tolerance is a plant's ability to persist in the understory at low light levels [38]. F. sylvatica in particular is considered to be one of, if not the most, shade tolerant species among important deciduous tree species in Germany [30,39]. F. sylvatica seedlings in this survey, however, responded to low light levels with the highest amount mortality in this experiment, but as light increased mortality decreased (Figure 2). This outcome is in line with other findings [40][41][42][43]. The ability of F. sylvatica to survive under somewhat dense P. serotina shrub layers with low light levels in the dense understory was expected [26][27][28]. That the mortality of T. cordata did not respond along the light gradient was surprising. One growing season seems to be too short to detect this effect as the seedlings are exposed to totally different biotic (e.g., competition, parasites, grazing) and abiotic factors (e.g., climate, soil, light, water, nutrients) in the forest compared to the nursery. This outcome may have been caused by other factors than the influence of light such as soil composition or soil properties, since every tree species has its own optimum in terms of soil acidity or soil aridity for instance [30,39]. Additionally, an influence by the strong competitor P. serotina is conceivable. Competition can arise when resources (here: light) become more available [44] so as P. serotina could have had reacted to the presence of the underplanted seedlings.

No Detectable Biomass Increases after the Growing Season
The theory of plants' optimal biomass allocation, known as the 'functional equilibrium hypotheses' [45,46] or 'balanced growth hypothesis' [47] suggests that plants increase allocation to organs that best access limiting resources. For example, allocation to leaves is increased when plants are limited by light [30,48]. In high light environments, plants invest more biomass to the roots [49,50] by increasing transpiration [51]. Therefore, plants react under changed light regimes with different biomass allocation strategies depending on which environmental factors change [29,48].
In this survey, the comparisons of the seedling biomass (TDM, AGB, RM) before and after the vegetation period revealed no significant changes in either trees species (Table 2). This may have been due to the destructive harvest after only one vegetation season and an acclimation process that had just started. Acclimation by seedlings to changed environmental factors has been documented in other studies [49,52,53]. Since none of our tested dry masses showed biomass increases (Table 2), the degree of acclimation of the small trees in this survey could not be clarified. Indeed, seedlings sprouts in the spring from the stock obtained from the previous year, but in our case the planting shock seemed to have strongly influenced the seedlings as they did not grow significantly during the vegetation period 2017. The seedlings' growth performance did not reveal significant differences in terms of the mechanical treatments applied to P. serotina.
However, the high survival rates of both F. sylvatica and T. cordata, in between the established population of P. serotina point towards the potential of underplanting. This could be a promising silvicultural alternative to outcompete the invasive species in the long run [26][27][28]. A retreatment of regrown P. serotina sprouts (following the initial mechanical treatments) after two to four vegetation periods would most likely strongly increase the probability of this approach to work. However, the impact of browsing should not be underestimated. If game populations are high, fencing might be necessary because native species are preferably browsed on. In addition to the expansive mechanical treatments, this would increase the costs strongly.
Unexpectedly, F. sylvatica LMF decreased in low light conditions in this survey ( Figure 3, Table 3). The effect size was minimal. This result also contradicts findings from other studies that support the theory of allocation, namely that LMF increases with shade in shade tolerant species [40,49,55]. Nevertheless, another study [41] also found an unexpected decrease in F. sylvatica LMF with shade. The authors there considered a fungal disease responsible for their results.

Shade Tolerant Species-Specific Reaction in Leaves
Results of SLA measurements in this survey confirmed findings of others that SLA decreases with increasing light availability [40,41,49,57]. The mean leaf area did not change along the light gradient, indicating that changes in SLA could be attributed to morphological changes in lamina thickness [49,58,59] and as morphological reactions of leaves to varying light conditions [29]. We found on average smaller SLA in F. sylvatica as compared to T. cordata along the entire gradient. Our model for T. cordata suggests a more sensitive reaction to changes in light availability in juvenile T. cordata as compared to juvenile F. sylvatica (Figure 4). Both species reacted to changed light availability resulting from mechanical treatments to P. serotina, which suggests an ongoing acclimation to the environmental conditions in the forest of Linde.

Study Area
The study area is located in Linde and belongs to the Zwillenberg-Tietz Foundation. Linde is located in eastern Germany (52 • 33'4.337" N, 12 • 40'55.08" E) inside the federal state of Brandenburg 50 km from the German capital Berlin (Figure 5a). The region around the study site is dominated by managed forests and agricultural land. Geologically characterized by flat-wave ground moraine plates and dune fields, the forest sites are rather nutrient poor. The climate of the region is classified as temperate with high temperatures in summer and moderately cold temperatures in winter. The annual average temperature is about 9 • C. Most rainfall occurs in summer during the vegetation period, with an annual precipitation of 520-572 mm [60]. The annual average temperature in 2017 was about 10 • C in the federal state of Brandenburg, which was 1. The study sites were placed in relatively pure and even-aged pine stands with a P. serotina understory (b2, c2, c3; Figure 5b). The stands differ by thinning regime. Stand b2 was last thinned in March 2017, prior to the planting activities. Stand c3 was unthinned and stand c2 was thinned before 2006. The abundance of P. serotina varied slightly throughout the stands. The stand situations encountered in Linde are representative of many pine forests in eastern Germany, where P. serotina grows invasively (Figure 5c).

Study Design
The study design was based on a block design. One block contained 18 individual 100 m 2 plots with randomized treatments. The plots were arranged with 5 m spacing to one another and a minimum distance of 2 m to existing skidding trails. Eleven block replicates were established for a total of 197 plots (one plot was excluded), of which 90 plots were in b2, 57 in c2, and 50 in c3 ( Figure  5b). The number of plots within the three stands differed due to the size of the stands. Each plot consisted of (1) a mechanical treatment applied to the P. serotina trees (felling or girdling), or control (without mechanical treatment), and (2) one out of two underplanted native tree species, or control (without underplanting). These plot-combinations existed fenced and unfenced. The protected plots were fenced with 1.80 m high game-proof wire to exclude grazing mammals, mainly ungulates. The plots were randomly spaced within the blocks (Figure 6a).

Study Design
The study design was based on a block design. One block contained 18 individual 100 m 2 plots with randomized treatments. The plots were arranged with 5 m spacing to one another and a minimum distance of 2 m to existing skidding trails. Eleven block replicates were established for a total of 197 plots (one plot was excluded), of which 90 plots were in b2, 57 in c2, and 50 in c3 (Figure 5b). The number of plots within the three stands differed due to the size of the stands. Each plot consisted of (1) a mechanical treatment applied to the P. serotina trees (felling or girdling), or control (without mechanical treatment), and (2) one out of two underplanted native tree species, or control (without underplanting). These plot-combinations existed fenced and unfenced. The protected plots were fenced with 1.80 m high game-proof wire to exclude grazing mammals, mainly ungulates. The plots were randomly spaced within the blocks (Figure 6a).
High shade tolerance in juvenile stages was required for the successful establishment of species under P. serotina [23][24][25][26][27][28]. Highly shade tolerant tree species native to Germany are Fagus sylvatica and T. cordata [30,37,39,56], therefore these two species were chosen. As planting material, 2-year-old wildstock seedlings of F. sylvatica and T. cordata were used. F. sylvatica seedlings had an initial biomass of 26.68 ± 11.02 g. The average biomass of T. cordata was 7.74 ± 4.12 g. The provenance of the planting material was Mark-Lusatia flatland for F. sylvatica (81005) and central and east German low and highlands for T. cordata (82303) [61]. Planting was conducted over 3 days in spring 2017. A total of 4189 seedlings (2109 F. sylvatica and 2080 T. cordata) were planted using a planting auger. Before the seedlings were planted, they were dipped into a hydrogel solution (GEFA Wurzelschutzgel) to improve their water storage and absorbance capabilities and to improve their overall survival rates [31]. Plant spacing was 2 × 1 m (5000 plants·ha −1 ), with 32 plants per plot (eight individuals × four rows, Figure 6b).  . sylvatica and 2080 T. cordata) were planted using a planting auger. Before the seedlings were planted, they were dipped into a hydrogel solution (GEFA Wurzelschutzgel) to improve their water storage and absorbance capabilities and to improve their overall survival rates [31]. Plant spacing was 2 × 1 m (5000 plants·ha −1 ), with 32 plants per plot (eight individuals × four rows, Figure 6b). As mechanical treatments to control P. serotina, felling and girdling were applied. Before the treatments were applied to the trees, their dbh (diameter in breast height) was measured using a calliper. This study focuses on trees with a diameter in breast height ≥ 5 cm. Smaller trees (dbh < 5 cm) were counted and mechanically treated in the same way. On average, 12.53 ± 11.94 P. serotina trees with a dbh ≥ 5 cm grew on each plot. A total of 403 P. serotina trees with a dbh ≥ 5 cm were felled and removed from the plots in March and April 2017. Girdling work was done by completely removing the bark and cambium on a 15 cm wide stem circumference section beneath the lowest living branch using a chain girdler. The girdling treatment was applied in August 2017, assuming this would have the greatest detrimental effect to suppress greater sprout responses at the end of the vegetation period [62]. A total of 363 P. serotina trees with a dbh ≥ 5 cm were girdled.

Seedling Delivery
We received 4247 2-year-old wildstock seedlings (2139 F. sylvatica and 2108 T. cordata) from the nursery and removed 58 before planting as a basic data set (nursery plants). These seedlings (30 F. sylvatica and 28 T. cordata) were randomly chosen from the delivered plants to record the initial conditions and mass fractions for later comparisons. Overall, 4189 seedlings (2109 F. sylvatica and 2080 T. cordata) were planted of which one F. sylvatica and 18 T. cordata seedlings died directly after As mechanical treatments to control P. serotina, felling and girdling were applied. Before the treatments were applied to the trees, their dbh (diameter in breast height) was measured using a calliper. This study focuses on trees with a diameter in breast height ≥ 5 cm. Smaller trees (dbh < 5 cm) were counted and mechanically treated in the same way. On average, 12.53 ± 11.94 P. serotina trees with a dbh ≥ 5 cm grew on each plot. A total of 403 P. serotina trees with a dbh ≥ 5 cm were felled and removed from the plots in March and April 2017. Girdling work was done by completely removing the bark and cambium on a 15 cm wide stem circumference section beneath the lowest living branch using a chain girdler. The girdling treatment was applied in August 2017, assuming this would have the greatest detrimental effect to suppress greater sprout responses at the end of the vegetation period [62]. A total of 363 P. serotina trees with a dbh ≥ 5 cm were girdled.

Seedling Delivery
We received 4247 2-year-old wildstock seedlings (2139 F. sylvatica and 2108 T. cordata) from the nursery and removed 58 before planting as a basic data set (nursery plants). These seedlings (30 F. sylvatica and 28 T. cordata) were randomly chosen from the delivered plants to record the initial conditions and mass fractions for later comparisons. Overall, 4189 seedlings (2109 F. sylvatica and 2080 T. cordata) were planted of which one F. sylvatica and 18 T. cordata seedlings died directly after planting activities. Therefore, the experiment started with 4170 seedlings (2108 F. sylvatica and 2062 T. cordata).

Seedling Mortality
The vitality of each seedling was recorded at the end of the vegetation period. Plants were considered as dead if they were either totally dried-out, had been browsed down to the ground, or were simply missing. Mortality was calculated on each plot as the ratio between dead and planted seedlings.

Seedling Harvest
Three randomly selected seedlings were destructively harvested (above-and belowground biomass, maximum two seedlings per planting row) on each plot at the end of the growing season 2017, before the seedlings shed their leaves. This resulted in a total of 393 harvested seedlings, 198 F. sylvatica and 195 T. cordata (three harvested seedlings on 12 underplanted plots in 11 replicates minus three seedlings, because one plot was excluded).

Light Measurements
All light measurements were carried out during the second week of July 2017. A single measurement was recorded on each plot using a hemispherical photography device equipped with a fisheye-lens (Behling, Solariscope SOL300). Measurements were conducted at 2 m above ground. In this study the ISF (indirect site factor) was used to classify the light environment on each plot. ISF is defined as proportion of diffuse radiation to open field condition, that reaches the measured point (here: per plot).

Lab Data Collection
Like the 58 removed nursery plants before, all 393 harvested seedlings were partitioned into leaves, branches, stems, and roots. These compartments were dried for 3 days in a temperature-controlled oven at 70 • C and weighed. Seedlings total dry mass (TDM) was the summed weight of all components: leaf mass (LM), branch mass (BM), stem mass (SM), and root mass (RM) (Equation (1)).
From the weight of the compartments, mass fractions could then be derived for every seedling. These were leaf mass fraction (LMF), branch mass fraction (BMF), stem mass fraction (SMF), and root mass fraction (RMF) (Equation (3)).
Subsamples of 20 fresh leaves of F. sylvatica and 10 fresh leaves of T. cordata were randomly selected out of a collection tank after all leaves had been fully removed from each harvested seedling. The subsamples were scanned to determine mean leaf area (LA, Equation (4)) and specific leaf area (SLA, Equation (5)

Data Analysis and Modelling
Statistical data analyses and graphs were processed using the statistical free software environment RStudio for R (Version 1.1.383 for Mac) [63]. Significance-level was p < 0.05.
Normality of distribution was checked using the Shapiro-Wilk test. Distribution was verified additionally in quantile-quantile-plots or frequency-plots. Homoscedasticity within groups (fencing, tree-species) was tested with Bartlett's test. The non-parametric Kruskal-Wallis test (one-way ANOVA of ranks) or ANOVA were used to analyze differences in groups depending on data distribution and homoscedasticity.
Correlations between non-normal distributed data were calculated by using the non-parametric method of Kendall's rank correlation coefficient tau (τ) to test for relationships between response variables (mortality rates, mass fractions, specific leaf area) and explanatory variables (ISF). Calculated means were compared using the independent two-sample t-test or Welch's test, depending whether or not the variances for the tested groups were equal. Chi-squared test was performed to test attributes (alive or dead seedlings) for independence within groups (fencing, species).
For modelling the mass fractions along the light gradient, generalized linear models (GLM) were used with indirect site factor (ISF) as independent variable (Equation (6)), where y was the vector of the dependent variable, x the predictors matrix, α the vector of weighed predictors and ε the vector of residuals.
The coefficient of determination (R 2 ) is defined as the proportion of the variance in the dependent variable (TDM, AGB, compartment mass fractions, SLA) which can be predicted from the independent variable (light, measured as ISF) (Equation (7)). The coefficient of determination was used to evaluate the goodness of fit of the calculated GLMs.
residual sum of squares total sum of squares (7) Generalized additive models (GAM) (Equation (8)) were used to estimate mortality along the light gradient (ISF), in order to determine an unbiased detection of trends in the data [64]. Mortality was modelled as function of available light, f as effect-specific regression spline function of the explanatory variable (ISF), α as intercept, and ε the error term assumed to be normally distributed with a mean value of zero and with finite variance (R-package 'mgcv') [65].

Conclusions
Since the results of this study are based on just 1 year of observation the following conclusions need to be checked in the near future. Surprisingly, we did not find higher mortality under low light conditions in T. cordata than in F. sylvatica, which is considered to be more shade tolerant than T. cordata. However, we could show that F. sylvatica and T. cordata show a relatively low mortality if underplanted in Scots pine stands that had been invaded by P. serotina. This finding shows the potential of this approach to grant seedlings access to light by using mechanical control against P. serotina. Retreatments of regrown sprout might be necessary to keep up the seedling's vitality. Fencing increases costs strongly. With one exception no changes in biomass allocation with decreasing light availability could be observed after one growing season. However, stronger morphological responses to the changed light environments as result of the treatments in both F. sylvatica and T. cordata can be expected in the up-coming years, as a result of a possibly delayed response after transplanting. The different light regimes on the plots will then most likely cause more noticeable differences in biomass allocation and growth, which will allow to identify best planting conditions. Therefore, the experiment in Linde is pursued and more data will continue to improve the recommendations that can be derived from this study for foresters dealing with the invasive tree species P. serotina.

Supplementary Materials:
The following are available online at http://www.mdpi.com/2223-7747/9/6/677/s1, Table S1: Seedling mortality, Table S2: Seedling data, Table S3 Acknowledgments: Gratitude is owed to the Zwillenberg-Tietz foundation for funding this study and giving us access to their stands for setting up this field experiment. We are grateful for receiving access to stands together with Martin Hasselbach to find suitable stands and also his support regarding information about the stands. We thank Marcus Wicke for his help in the field. We would also like to thank Ulrike Westphal for supporting fieldwork, data acquisitions in the field, performing the lab work, and her constructive comments regarding plot measuring and tree labeling. Many thanks go to Michael Unger and Ray Wollenzien for felling and removing the trees in stand b2 in only one day and Steffi Heinrichs for her data acquisition concerning the vegetation. Thanks are also due to our student assistants for their help with setting up the field experiment, labeling, data acquisition, felling and clearing plots and performing lab work. We would like to thank Christian Ammer for guidance and helpful comments on an earlier draft of this manuscript. We further acknowledge the linguistic corrections made by Kathleen Regan (USA). We would like to thank the reviewers for their helpful comments towards improving our manuscript.