Early Post-Germination Physiological Traits of Oak Species Under Various Environmental Conditions in Oak Forests

Abstract

Early post-germination physiological responses determine oak seedling establishment success under changing environmental conditions. This study investigated four oak species (Quercus cerris, Q. frainetto, Q. petraea, and Q. pubescens) through direct seeding experiments across four locations in Serbia representing varying habitat conditions. Physiological parameters (quantum yield of photosystem II, total stomatal conductance, and leaf vapor pressure deficit) were measured intensively during the first growing season, along with morphological traits and survival rates. Results revealed that microclimatic and soil conditions exerted stronger effects on seedling physiology than species identity, with air humidity, temperature, and soil moisture being primary drivers of photosynthetic performance. Surviving seedlings exhibited 18% higher PhiPSII and 128% higher stomatal conductance compared to non-survivors, demonstrating that physiological performance is a reliable predictor of establishment success. Species-specific differences were evident. Q. cerris and Q. frainetto maintained the highest photosynthetic activity across sites, Q. pubescens showed intermediate resilience, and Q. petraea displayed greater sensitivity to environmental stress. These findings highlight the dominant role of microsite conditions in shaping early seedling physiology and survival. Physiological measurements, particularly PhiPSII and gtw, provide useful early indicators of establishment success during the first growing season following direct seeding.

1. Introduction

In Europe, oak forests have immeasurable economic and ecological importance. Oak forests provide valuable wood products, regulate hydrological and biogeochemical cycles, support high biodiversity levels, and serve numerous other important functions [1]. In Europe, oak communities are clearly differentiated due to diverse environmental conditions and the adaptability of these species [2,3,4].

However, European oak forests are undergoing major structural and functional changes driven by climate change, various abiotic and biotic stressors, and challenges related to natural regeneration and sustainable management [5,6,7]. There are views that the sustainability of these ecosystems may depend on future management strategies that consider both strategic (long-term) elements, such as adaptation to climate change and future disturbances [1,8,9], and operational (short-term) elements, including adjustments of silvicultural treatments, particularly those related to natural regeneration and early interventions in the initial post-regeneration developmental phase [10,11,12].

Oak species occupy diverse ecological niches and display variable responses to drought, shade, and soil conditions. For example, Quercus cerris and Quercus frainetto typically occupy drier habitats at lower elevations within the hilly zone [13,14,15], whereas Q. petraea, although capable of mixing with these species, is generally found at higher elevations under more humid conditions [7,13,16]. Finally, Q. pubescens is a xerophilous and thermophilous species that occupies the warmest and driest habitats, yet it also has the capacity to mix with the previously mentioned oaks [13,17]. Variations in oak responses to changing environmental conditions can be quantitatively evaluated using key physiological performance indicators, such as photosynthetic and respiratory rates, stomatal conductance, and water-use efficiency [18]. Understanding the physiological basis of these interspecies differences is essential for predicting regeneration success and for guiding silvicultural decisions under changing climatic conditions [19].

Early post-germination development represents a critical phase in the life cycle of tree species, during which seedlings establish functional photosynthetic and root systems, regulate water balance, acclimate to microenvironmental conditions, and thereby achieve their initial growth parameters [20,21,22]. The developmental phase preceding and associated with seed germination also influences the characteristics of early post-germination development, as the abundance, survival, and initial growth of oak seedlings depend strongly on acorn traits, intense herbivory presence (e.g., small mammals), and habitat conditions [23,24,25]. During early post-germination development, physiological mechanisms, such as chlorophyll fluorescence responses, stomatal regulation, maintenance of water potential, and antioxidant defenses, play a decisive role in seedling survival [26,27]. In this context, oaks can exhibit markedly different responses under varying environmental conditions, which reflects the adaptive traits of these tree species [22,28,29].

The profound and accelerating changes in habitat conditions in which oaks occur, coupled with increasingly intense competition from co-occurring species that suppress oak regeneration and growth, necessitate continuous adaptation and adjustment of their physiological and ecological responses [30]. Adaptation in oak species, through modification of physiological responses, is a fundamental mechanism supporting their resilience and long-term persistence [31]. Various morpho-physiological traits related to water retention have been used among different oak species [32,33,34]. Equally important are findings highlighting oaks’ high adaptive capacity to heat stress, particularly through regulation of stomatal conductance and enhanced antioxidant activity [10,35,36]. Oaks can adjust the rates of photosynthesis, transpiration, and respiration in response to water availability, temperature, and light intensity, thereby optimizing energy balance and resource-use efficiency [37]. Furthermore, the enhancement of the adaptive capacity of oak species could result from high interspecific hybridization rates [38,39].

The physiological responses of oaks may be highly significant, particularly during the early stages of development, given the frequent challenges associated with regenerating these forests. This issue is especially pronounced in Southeastern Europe, where natural regeneration methods are predominantly used in oak stands and require adaptive management strategies that account for species-specific ecological amplitudes and physiological tolerances [7,13]. The regeneration of oak forests, both in this region and more broadly, is predominantly carried out using various variants of the shelterwood system, with the early post-germination developmental phase typically assessed based on seedling morphological characteristics [11,40,41].

Although the physiological processes of oak seedlings during the early post-germination phase have been extensively studied in controlled nursery settings, there has been significantly less research on their physiological performance in natural forest environments. In these settings, fluctuations in microclimate factors and interactions with other organisms create complex challenges that require adaptive responses. Unlike most prior work, this study combines high-frequency physiological measurements (collected repeatedly throughout the first growing season) with survival outcomes, explicitly linking early physiological performance to seedling survival. This multi-site, multi-species approach, under natural environmental variability, provides novel insights into the mechanisms underlying the success of oak restoration in heterogeneous forest environments. Based on that, and aiming to separate environmental filtering effects from intrinsic species-level physiological differences, the following hypotheses were formulated in this study:

H1: 

Variation in habitat conditions is associated with differences in the physiological parameters of oak seedlings during the early post-germination developmental stage, with location exerting stronger effects than species identity.

H2: 

Physiological responses vary among oak species during early post-germination development, with species showing distinct responses to environmental variation.

H3: 

Seedlings with higher physiological performance exhibit greater survival during early post-germination growth.

2. Materials and Methods

2.1. Seed Collecting

Multiple-criteria decision analysis (MCDA) was performed to select locations for acorn collecting (five locations for every oak species, Figure 1). The MCDA included three criterion groups.

(1) Biodiversity attributes: oak species diversity, geographical distribution, and genetic diversity. These variables ensured that selected provenances captured the natural range of variation within each oak species.

(2) Ecological and stand-level attributes: complexity of ecological conditions (soil type, microclimate, topography), overall forest condition/preservation status, dominant forest origin (natural vs. artificially regenerated stands), and vegetation complexity. These attributes ensured that the source stands represented typical and ecologically relevant habitats for each species.

(3) Regeneration dynamics: urgency of regeneration needs (based on management plans) and complexity of natural regeneration conditions. These factors ensured that the selected locations reflected current silvicultural challenges related to oak regeneration in Serbia.

Each attribute was scored on a standardized scale following expert assessment, and all criteria groups were assigned equal weight. Provenances with the highest composite MCDA scores were selected for acorn collection.

Seed was collected during September and October 2024. At each selected provenance, between 10 and 20 mother trees per species were sampled, spaced at least 50 m apart to reduce the likelihood of relatedness. From each tree, approximately 1000 mature acorns were collected and harvested randomly. The collected acorns were submerged in water for 30 min. After that, floating acorns were removed from water and discarded, while the rest were stored at the University of Belgrade—Faculty of Forestry in the refrigerator at a temperature of 4 °C (±2 °C). During storage, acorns were sprinkled with water to maintain moisture.

2.2. Experiment Plots

Research was conducted across 4 locations in Serbia: Cer, Debeli Lug, Vraćevšnica, and Žiča (Figure 1,

Table 1. Experimental plots data.

Location	Coordinates	Plant Community	Altitude (m)	Parent Material	Soil Type	Organic Layer (cm)	Canopy Closure (%)	Overstory Species
Cer	44.64881453° N, 19.42372202° E	Quercetum frainetto-cerridis Rudski (1940) 1949	217	Sandstones, argillophyllites and phyllites	Luvisol	5	70	Quercus frainetto, Tilia cordata, Carpinus betulus
Debeli Lug	44.32694° N, 21.88722° E	Festuco drymeiae-Quercetum petraeae Janković 1968	537	Schists	Cambisol	4/5	70	Quercus petraea, Carpinus betulus, Fraxinus excelsior, Fagus sylvatica
Vraćevšnica	44.06519794° N, 20.60490133° E	Quercetum frainetto-cerridis Rudski (1940) 1949	430	Sandstones	Cambisol	7	70	Quercus frainetto, Quercus cerris, Fagus sylvatica
Žiča	43.6925° N, 20.64333° E	Quercetum frainetto-cerridis Rudski (1940) 1949	241	Clay	Planosol	4	80	Quercus frainetto, Quercus cerris, Carpinus betulus

). These locations were selected to represent a gradient of environmental conditions typical of hilly oak forest ecosystems in Serbia. The locations differ in altitude, soil type, moisture availability, and microclimatic regimes, thereby capturing the environmental contrasts that influence oak seedling establishment in the region. This selection ensured that the experiment encompassed the realistic range of habitat conditions under which oak regeneration typically occurs. Two experimental plots, each 20 × 10 m, were established at each location. Ecological conditions and structural characteristics of the stand were defined on the experimental plots.

To assess canopy cover, hemispherical photographs were taken across all experimental plots using a Nikon Coolpix 5000 digital camera (Nikon Corp., Tokyo, Japan) fitted with an FC-E8 hemispherical (“fish-eye” or open-sky) lens. Image acquisition was conducted on clear, cloud-free days between 9:00 AM and 12:00 PM to ensure consistent lighting conditions. The camera was positioned 1.30 m above the ground and carefully leveled both horizontally and vertically during each capture. The obtained images were subsequently analyzed using the Gap Light Analyzer (GLA) software (version 2.0) [42]. In the middle of each experimental plot, a data logger (LOG210, Dostmann Electronic, Wertheim, Germany) for recording air temperature, humidity, and dew point temperature was installed on a tree, approximately 80 cm above the ground, facing north and protected from precipitation. The devices were set to record data every 30 min. Additionally, current soil moisture and temperature were measured using the WET150 Sensor (Delta-T Devices, Cambridge, UK) from the WET150 Kit at a depth of 5 cm. These parameters were measured at each corner and at the center of the experimental plots, and the average values were used. Although microclimate can vary within plots, this standardized design enabled consistent comparisons of relative differences among locations.

2.3. Applied Method—Direct Seeding

At each location, two separate experimental plots were established within the selected area, providing replication at the plot level. In November 2024, experimental plots were prepared immediately before direct seeding, including stand and soil preparation. After preparation, each experimental plot was divided into more minor equal subplots (1 × 1 m). Plots were arranged in a randomized complete block design with species and provenance as experimental factors within plots. Each species–provenance combination was assigned to a different 1 × 1 m subplot in each of the two plots, ensuring no repetition of the same treatment within a single plot. Every subplot was divided into five rows equally distant from each other (Figure S1). All seeds were sown in prepared rows at approximately 3 cm depth and covered with the same soil, resulting in a density of 50 acorns per m² per subplot.

2.4. Morphological and Physiological Measurements

At the end of the study period, the number of surviving seedlings per experimental plot was recorded. Survival rate was determined as the percentage of oak seedlings that remained alive at the end of the study period. The sample size was set at five plants per subplot, with replication. Selected oak seedlings were measured using a digital caliper (diameter at ground level with accuracy ±0.1 mm) and a measuring tape (height with accuracy ± 0.1 cm) at the start and the end of the study period. Physiological measurements were conducted in summer 2025 using a LI-600 Porometer/Fluorometer (LI-COR Biosciences, Lincoln, NE, USA). At all locations intensive measurements were performed, every week for 4 weeks in a row (during June and July) seedlings’ youngest fully expanded leaf was measured in the morning (AM dataset, from 08:00 h to 14:00 h) and in the afternoon (PM dataset, from 15:00 h to 18:00 h), and in August and September were carried out for one week each. Measurements were taken weekly in June and July because seedlings had only recently emerged and were highly sensitive to rapid changes in temperature, moisture, and atmospheric demand. Frequent sampling during this phase enabled the capture of early-season fluctuations when stress exposure and mortality risk were greatest. In contrast, August and September documented broader late-season patterns. Environmental conditions during this period changed more gradually, and surviving seedlings either stabilized or recovered following early-season stress. Accordingly, monthly measurements provided sufficient temporal resolution for capturing these late-season trends. Physiological parameters that were recorded were the quantum yield of photosystem II (PhiPSII as (Fm′ − Fs)/Fm′, where Fs is minimum fluorescence in light and Fm′ is maximum fluorescence in light), total stomatal conductance to water vapor (gtw), and leaf vapor pressure deficit (VPDleaf).

2.5. Statistical Analysis

All analyses were conducted in R version 4.3.2 (R Core Team, 2024 [43]). Environmental variables (air temperature, air relative humidity, soil temperature, and soil moisture content) were summarized as location-level daily means across the study period and used descriptively to characterize microclimatic differences among locations. Although soil temperature and soil moisture were measured at five points within each plot, these values were averaged to obtain a single plot-level estimate. However, in the linear mixed-effects models, the same environmental parameters were included as time-varying covariates, representing measurement-time conditions rather than experimental treatments, thereby accounting for temporal variation in microclimate during physiological assessments. Descriptive statistics were computed for each species and location. Data normality was verified using the Anderson–Darling test (nortest [44]), while homogeneity of variances was tested using Levene’s test (car [45]).

Morphological parameters (seedlings’ height and diameter) were analyzed using two-way ANOVA (aov, stats package, R version 4.3.2, R Core Team, 2024) with species and location as fixed factors, followed by Tukey’s HSD (multcomp [46]) tests (p < 0.05) to compare species within locations.

Physiological parameters (PhiPSII, gtw, and VPDleaf) were measured repeatedly on the same seedlings across multiple dates. Because repeated observations violate the independence assumptions of ANOVA, these traits were analyzed using linear mixed-effects models (LMEs) as the primary inferential framework. Separate models were fitted for survivors and non-survivors, and for AM and PM measurement periods, reflecting the structure used in the Section 3. This separation was necessary because mortality during the study was non-random, and surviving seedlings exhibited different physiological characteristics than those that died, making pooled analyses statistically inappropriate. This post hoc grouping by survival status represents a deliberate trade-off between biological realism and strict inferential independence, prioritizing ecologically meaningful interpretation of physiological differences.

Models included species, location, and their interaction as fixed effects. To account for environmental conditions at the time of measurement, soil moisture content, soil temperature, air temperature, and air relative humidity were included as time-varying covariates (i.e., reflecting temporal microclimatic variation during physiological measurements). A random intercept for Seedling_ID (the same seedlings measured repeatedly) was included to model within-individual correlation.

Provenance was initially considered as a potential explanatory factor; however, it was removed from the fixed-effects structure for two methodological reasons. First, provenance was highly unbalanced across locations and species, and several provenances had very few surviving seedlings. This lack of replication prevented reliable estimation of provenance-level effects and led to single-model fits. Second, the study’s primary aims focused on species-level and habitat-level variation, whereas provenance served only as a source of random biological variation rather than as a factor of experimental interest. For these reasons, provenance was not included as a fixed factor in the LME models (fitted using lme4 [47] and lmerTest [48]).

Two-way ANOVA (aov, stats package, R version 4.3.2, R Core Team, 2024) was additionally applied to the physiological dataset, separately for survivors and non-survivors and for AM and PM measurements, to assess species, location, and species × location differences within each subset. These ANOVAs were based on raw physiological observations within each group and served as complementary descriptive analyses to support the interpretation and visualization of species × location patterns. Because these models do not account for the repeated measures, all inferential conclusions regarding physiological responses and environmental effects rely exclusively on the LME models.

To examine how physiological responses varied across environmental gradients, continuous environmental variables (air temperature, air relative humidity, soil temperature, soil moisture content) and VPDleaf were grouped into classes. For each variable, the full range of values recorded during the study period was divided into quartiles, and each quartile was assigned to one of four categories (Class 1–Class 4). This approach provided a standardized method for visualizing physiological responses across comparable environmental conditions without imposing arbitrary thresholds. These classes were used only for descriptive visualization in the interaction plots and do not represent statistical groupings.

A binomial logistic regression model was fitted with survival status (alive at the end of the season = 1, dead = 0) as the response variable to evaluate whether physiological traits predicted seedling survival. Predictor variables included seedling-level mean PhiPSII and mean gtw (averaged across all measurement dates), species, and location. Model coefficients were exponentiated to obtain odds ratios, representing the multiplicative change in the odds of survival per unit increase in each predictor. Model performance was evaluated by calculating overall classification accuracy, sensitivity (true-positive rate), and specificity (true-negative rate) using a probability threshold of 0.5. A confusion matrix was used to quantify the number of true positives, true negatives, false positives, and false negatives. To facilitate interpretation of odds ratios, PhiPSII was scaled by dividing by 0.05, such that odds ratios represent the change in survival odds per 0.05-unit increase (approximately 5% of the typical range). Mean gtw values were converted from mol m⁻² s⁻¹ to mmol m⁻² s⁻¹ (multiplied by 1000), such that odds ratios represent the change per 1 mmol m⁻² s⁻¹ increase in stomatal conductance. Variance inflation factors (VIF) were calculated to assess multicollinearity among predictors, with all VIF values < 2 indicating no substantial collinearity issues.

Figures were generated in ggplot2 [49] and arranged using patchwork [50], with a consistent color scheme applied across all figures (Q. cerris-red, Q. frainetto-green, Q. petraea-blue, Q. pubescens-purple) to facilitate visual comparison across analyses. Data manipulation and summarization were performed with dplyr [51] and lubridate [52] packages.

3. Results

3.1. Environmental Conditions

Air temperature (A) and relative humidity (B) showed similar patterns across locations, with values ranging from 8.6–36.4 °C and 13.97%–100%, respectively, and with a high day-to-day variability (Figure 2). Soil temperature (°C) peaked in late June–early July at all locations (approximately 30–40 °C), but the magnitude and subsequent trends differed among locations. Debeli Lug exhibited the highest mid-summer soil temperatures and the most significant variability, followed by Cer. In contrast, Vraćevšnica showed intermediate values, and Žiča generally maintained the coolest soil conditions over the study period. In contrast, soil moisture content (D) varied noticeably across locations, establishing a transparent moisture gradient. Soil moisture was consistently highest at Debeli Lug (mean: 14.6% ± 5.5%) and Žiča (13.1% ± 4.5%), intermediate at Cer (7.2% ± 3.9%), and lowest at Vraćevšnica (5.7% ± 3.6%). Debeli Lug showed increasing soil moisture content in late September, while Žiča and Cer exhibited a declining trend.

3.2. Survival

Survival rates varied substantially among planting locations (Table S1, Figure 3). Vraćevšnica had the highest survival (91.3%, 723 of 792 seedlings), followed by Debeli Lug (81.2%, 739 of 910 seedlings). In contrast, lower survival rates were at Cer at 71.1% (118 of 166 seedlings) and Žiča at 70.9% (95 of 134 seedlings), representing approximately 20%–30% higher mortality compared to Debeli Lug and Vraćevšnica, which even had an initial higher number of seedlings. The highest survival rate, when analyzed across species, was Q. frainetto (89.6%, 493 of 550 seedlings), followed by Q. cerris (87.0%, 456 of 524), Q. pubescens (81.0%, 346 of 427), and Q. petraea (75.9%, 380 of 501). The lowest survival was observed for Q. petraea at Cer (55.6%, 20 of 39 seedlings), whereas the highest survival was detected for Q. cerris and Q. frainetto at Vraćevšnica (97.9% and 97.3%, respectively).

3.3. Morphology

Mean seedlings’ height ranged from 5.58 ± 1.90 cm (Q. pubescens at Vraćevšnica) to 8.54 ± 3.75 cm (Q. cerris at Žiča), and mean diameter varied between 1.23 ± 0.57 mm (Q. pubescens at Cer) and 2.26 ± 0.42 mm (Q. cerris, Debeli Lug) (Table S2, Figure 4). Across locations, seedlings at Debeli Lug and Vraćevšnica had larger diameters, but had lower heights than those at Cer and Žiča. Significant differences were detected in both height and diameter among oak species and across experimental locations (

Table 2. Two-way ANOVA results for morphological traits (height and stem diameter) of oak seedlings across species and locations (*** p < 0.001; ns—not significant).

Morphological Parameter	Source of Variation	df	F-Value	p-Value	Significance
Height (cm)	Species	3	10.38	1.14 × 10−6	***
	Location	3	21.80	1.95 × 10−13	***
	Species × Location	9	1.88	0.052	ns
Diameter (mm)	Species	3	37.47	2.90 × 10−22	***
	Location	3	47.53	1.16 × 10−27	***
	Species × Location	9	5.96	5.00 × 10−8	***

). ANOVA revealed strong effects of species (F = 10.38, p < 0.001) and location (F = 21.80, p < 0.001) on seedling height, while the interaction between species and location was marginal (F = 1.88, p = 0.052). However, diameter indicated apparent variation among species (F = 37.47, p < 0.001) and locations (F = 47.53, p < 0.001), with a highly significant interaction between the two (F = 5.96, p < 0.001). Tukey HSD post hoc test indicated that Q. cerris and Q. frainetto seedlings tended to be taller than Q. petraea and Q. pubescens (p < 0.001), whereas Q. cerris showed the greatest diameters, particularly at the Debeli Lug and Vraćevšnica (Table S2).

3.4. Physiology

Anderson–Darling and Levene’s tests were conducted to assess normality and homogeneity. Variance partitioning analysis of the main physiological parameters PhiPSII and gtw (Table S3) showed that provenance contributed minimally to the total variance (<3%) for all parameters, indicating that most variation occurred at the residual level. Slightly higher provenance effects were observed for PhiPSII in surviving seedlings, measured in PM (2.47%), whereas all other models were dominated by residual variance (>97%). Provenance was not included as a fixed effect because mortality and unbalanced representation across locations prevented meaningful estimation of provenance-level differences.

Analysis of AM measurements of the main physiological traits (PhiPSII and gtw) revealed that seedlings that survived throughout the study period exhibited 18.3% higher PhiPSII (p < 0.001) and 128.3% higher gtw (p < 0.001) than seedlings that died during the study (Figure 5). Because the sample size decreased over time (17.9% of seedlings were lost before September) and only survivors with higher physiological performance remained in the sample, it was justified to analyze the survival groups separately.

During the morning (AM) measurements, survivors showed generally high PhiPSII across species, with the highest values recorded at Žiča and the lowest at Vraćevšnica (Figure 6, Table S4). Although species differed slightly in their PhiPSII values, all four species followed the same location pattern. Individual PhiPSII values ranged widely, but this variation was captured in the linear mixed-effects models and reflected the strong influence of location conditions. In survivors, gtw displayed greater variability than PhiPSII and showed apparent location-specific differences (Figure 7, Table S4). For most species, gtw was highest at either Vraćevšnica or Debeli Lug and lowest at Žiča. Species did not share a uniform ranking. For example, Q. frainetto reached its highest gtw at Debeli Lug, whereas Q. pubescens peaked at Cer. VPDleaf in survivors showed the expected spatial pattern, with generally higher values at Vraćevšnica and lower values at Cer, consistent with the environmental gradients described earlier. In the afternoon (PM), survivors showed slightly lower PhiPSII values than in the morning. Still, they retained the same overall location ranking, with Žiča maintaining the highest values and Vraćevšnica the lowest (Figure 6, Table S5). Patterns of gtw also remained location-dependent, with higher conductance generally observed at Debeli Lug and lower values at Cer or Žiča, indicating a consistent diurnal modulation of physiological activity rather than a shift in spatial trends.

In the morning (AM), non-surviving seedlings generally showed lower PhiPSII than surviving individuals, with apparent differences among locations (Figure 6, Table S6). For most species, PhiPSII was highest at Debeli Lug or Žiča and lowest at Vraćevšnica, although the magnitude of these differences varied among species. In non-survivors, gtw showed substantial variability across sites (Figure 7, Table S6). Several species exhibited their highest gtw at Vraćevšnica, while others peaked at Debeli Lug, and the lowest values for most species occurred at Žiča. These patterns were consistent with the broader physiological differences between survivors and non-survivors. During the afternoon (PM), PhiPSII in non-survivors declined slightly relative to the morning but maintained the same overall location-driven pattern (Figure 6, Table S7). Across species, higher PhiPSII values were generally observed at Cer or Žiča, whereas the lowest occurred at Vraćevšnica. PM gtw also followed clear site-level gradients, with the highest values occurring at either Vraćevšnica or Debeli Lug, depending on species, and consistently low values at Žiča (Figure 7, Table S7). Overall, non-survivors exhibited lower and more variable PhiPSII and gtw than survivors, and their physiological responses were more strongly structured by location than by species, consistent with the linear mixed-effects model results.

Linear mixed-effects (LME) models (

Table 3. Linear mixed effects model results for survivors (* p < 0.05, ** p < 0.01, *** p < 0.001).

Physiological Parameter	Term	AM			PM
		df	t-Value	p-Value	df	t-Value	p-Value
PhiPSII	(Intercept)	3499.9	17.78	9.5 × 10−68 ***	2959.3	15.61	7.99 × 10−53 ***
	Q. frainetto	632.2	−0.59	0.554	569.2	0.99	0.324
	Q. petraea	654.4	−2.01	0.045 *	592.0	0.12	0.903
	Q. pubescens	634.3	1.13	0.258	570.9	1.11	0.268
	Debeli Lug	750.1	1.87	0.062	706.7	2.26	0.024 *
	Vraćevšnica	643.8	−6.93	1.0 × 10−11 ***	554.5	0.03	0.974
	Žiča	600.5	2.42	0.016 *	558.8	4.19	3.18 × 10−5 ***
	Soil moisture content	3099.4	−2.29	0.022 *	2508.4	−5.31	1.21 × 10−7 ***
	Soil temperature	3030.8	−7.30	3.5 × 10−13 ***	2479.8	1.68	0.094
	Air temperature	3015.6	9.01	3.7 × 10−19 ***	2491.2	−1.49	0.137
	Air humidity	3013.2	8.22	3.1 × 10−16 ***	2463.1	5.02	5.67 × 10−7 ***
	Q. frainetto × Debeli Lug	585.0	0.34	0.732	536.3	−0.23	0.815
	Q. petraea × Debeli Lug	603.0	2.16	0.031 *	554.3	−1.02	0.308
	Q. pubescens × Debeli Lug	579.3	−0.61	0.544	532.5	−0.53	0.600
	Q. frainetto × Vraćevšnica	582.0	1.36	0.173	534.3	−0.74	0.461
	Q. petraea × Vraćevšnica	600.3	1.05	0.295	552.0	−0.93	0.353
	Q. pubescens × Vraćevšnica	575.0	−1.54	0.125	529.5	−0.80	0.424
	Q. frainetto × Žiča	564.3	0.74	0.457	522.0	−0.62	0.539
	Q. petraea × Žiča	548.6	1.83	0.067	513.0	−0.01	0.989
	Q. pubescens × Žiča	523.9	−0.21	0.834	493.0	−0.72	0.469
gtw	(Intercept)	3410.2	6.65	3.48 × 10−11 ***	2945.2	−1.94	0.052
	Q. frainetto	574.9	−1.30	0.193	611.7	0.72	0.469
	Q. petraea	593.8	−1.29	0.197	636.6	0.88	0.381
	Q. pubescens	576.3	0.57	0.568	613.6	1.37	0.171
	Debeli Lug	649.1	2.73	0.006 **	769.6	4.50	7.71 × 10−6 ***
	Vraćevšnica	582.2	−0.32	0.752	594.7	5.52	5.09 × 10−8 ***
	Žiča	555.6	−4.03	6.5 × 10−5 ***	599.6	0.03	0.974
	Soil moisture content	3061.3	−6.03	1.87 × 10−9 ***	2539.5	5.61	2.26 × 10−8 ***
	Soil temperature	3012.7	−0.30	0.762	2510.0	−3.78	1.63 × 10−4 ***
	Air temperature	3001.7	−7.06	2.05 × 10−12 ***	2521.5	3.70	2.21 × 10−4 ***
	Air humidity	3000.4	10.92	2.91 × 10−27 ***	2492.6	12.22	2.08 × 10−33 ***
	Q. frainetto × Debeli Lug	545.3	1.75	0.081	573.8	0.05	0.960
	Q. petraea × Debeli Lug	560.1	0.58	0.564	593.5	−1.23	0.221
	Q. pubescens × Debeli Lug	541.9	−0.60	0.548	569.4	−1.30	0.194
	Q. frainetto × Vraćevšnica	543.5	0.85	0.394	571.4	−1.91	0.057
	Q. petraea × Vraćevšnica	558.3	0.38	0.703	590.9	−2.50	0.013 *
	Q. pubescens × Vraćevšnica	539.1	−1.07	0.286	565.9	−2.64	0.008 **
	Q. frainetto × Žiča	533.3	1.13	0.258	557.0	−0.83	0.410
	Q. petraea × Žiča	524.3	0.89	0.373	546.5	−0.90	0.368
	Q. pubescens × Žiča	506.8	−0.51	0.610	523.9	−1.55	0.123

and

Table 4. Linear mixed effects model results for non-survivors (* p < 0.05, ** p < 0.01, *** p < 0.001).

Physio- Logical Parameter	Term	AM			PM
		df	t-Value	p-Value	df	t-Value	p-Value
PhiPSII	(Intercept)	439.6	0.22	0.825	292.0	0.52	0.603
	Q. frainetto	114.2	0.87	0.385	105.6	1.45	0.150
	Q. petraea	121.4	0.17	0.867	109.8	0.06	0.956
	Q. pubescens	101.6	0.42	0.679	88.9	1.27	0.206
	Debeli Lug	115.2	2.06	0.042 *	105.8	−0.55	0.585
	Vraćevšnica	102.7	0.20	0.840	94.6	0.88	0.383
	Žiča	119.6	4.21	4.940 × 10−5 ***	100.7	2.70	0.008 **
	Soil moisture content	385.4	−6.36	5.668 × 10−10 ***	252.6	−0.48	0.629
	Soil temperature	378.6	−1.87	0.063	257.5	−1.29	0.199
	Air temperature	370.6	3.47	6.0 × 10−4 ***	251.6	3.80	2.0 × 10−4 ***
	Air humidity	367.3	5.33	1.68 × 10−7 ***	246.9	−0.17	0.868
	Q. frainetto × Debeli Lug	87.1	−1.00	0.321	79.0	−0.65	0.517
	Q. petraea × Debeli Lug	99.3	0.44	0.661	89.4	0.78	0.436
	Q. frainetto × Vraćevšnica	74.7	0.03	0.973	70.1	−0.58	0.566
	Q. petraea × Vraćevšnica	90.3	0.12	0.906	82.3	−0.41	0.683
	Q. pubescens × Vraćevšnica	91.3	−0.29	0.769	81.1	−1.28	0.206
	Q. frainetto × Žiča	105.3	−2.28	0.025	97.4	−2.37	0.020 **
	Q. petraea × Žiča	102.5	−0.80	0.428	92.9	−0.71	0.481
	Q. pubescens × Žiča	94.9	−1.27	0.207	84.5	−1.76	0.082
gtw	(Intercept)	435.2	5.35	1.40 × 10−7 ***	311.7	3.21	0.001 **
	Q. frainetto	154.4	0.80	0.426	171.4	1.21	0.228
	Q. petraea	165.2	1.20	0.230	172.5	−0.37	0.709
	Q. pubescens	136.4	0.09	0.925	141.2	−0.66	0.508
	Debeli Lug	156.7	2.80	0.006 **	172.8	1.88	0.062
	Vraćevšnica	136.3	1.68	0.096	149.2	−0.82	0.414
	Žiča	163.6	0.46	0.645	164.4	−0.93	0.352
	Soil moisture content	411.4	−1.56	0.119	292.9	−2.06	0.040 *
	Soil temperature	405.2	−2.31	0.022 *	303.0	−3.74	2.19 × 10−4 ***
	Air temperature	395.8	−1.56	0.120	292.6	0.18	0.859
	Air humidity	392.7	−1.84	0.066	290.4	3.26	0.001 **
	Q. frainetto × Debeli Lug	111.2	−1.89	0.061	113.6	−1.18	0.239
	Q. petraea × Debeli Lug	129.9	−1.39	0.167	133.0	0.44	0.660
	Q. frainetto × Vraćevšnica	90.2	0.65	0.515	91.2	−1.01	0.317
	Q. petraea × Vraćevšnica	115.5	−0.02	0.987	117.5	−1.02	0.312
	Q. pubescens × Vraćevšnica	117.9	0.29	0.774	119.1	−0.49	0.622
	Q. frainetto × Žiča	139.8	−0.95	0.345	151.9	−0.89	0.373
	Q. petraea × Žiča	133.3	−1.24	0.217	135.4	−0.35	0.728
	Q. pubescens × Žiča	123.0	−0.24	0.814	124.2	0.60	0.548

) revealed that location had the most decisive influence on seedling physiological parameters, whereas species and their interactions contributed less prominently. For surviving seedlings, PhiPSII differed among species and was firmly structured by location in both AM and PM, indicating that species-specific photosynthetic responses depended on habitat conditions. PhiPSII in survivors was further shaped by microclimate: higher soil moisture and soil temperature were associated with lower PhiPSII, whereas higher air temperature and air humidity were associated with higher PhiPSII. Q. petraea generally showed slightly lower PhiPSII compared with Q. cerris, and survivors at Vraćevšnica had consistently lower PhiPSII than at other locations, while those at Žiča maintained the highest values. For non-surviving seedlings, LME indicated that location, rather than species, explained most of the variation in PhiPSII. Non-survivors at Žiča and Debeli Lug maintained relatively higher PhiPSII compared with those at Vraćevšnica and Cer, and soil moisture, air temperature, and air humidity again emerged as key drivers in the mixed-effects models.

For gtw, LME results confirmed the dominant role of location. In survivors, gtw was reduced under low soil moisture and low air temperature, and increased with higher air humidity. In PM measurements, soil moisture, air temperature, and air humidity all remained significant predictors. In non-survivors, gtw was strongly influenced by location and by shifts in air humidity and soil temperature, particularly in the afternoon. ANOVA (Table S8) detected similar location effects but did not capture the microclimatic influences revealed by the linear mixed-effects models. VPDleaf also exhibited strong site-level differences, with LME again identifying location as the primary driver. Species effects were not significant, indicating a dominant role of environmental variation.

Post hoc Tukey tests (Table S9) revealed species differences only for PhiPSII. Q. petraea maintained significantly lower PhiPSII than Q. cerris and Q. frainetto, while Q. pubescens showed intermediate values (supporting H2). However, the LME results indicate that environmental gradients explained more variation in physiological performance than species identity, supporting hypothesis H1.

The significant effects of environmental covariates in the LME models (Table 3 and Table 4) demonstrate that short-term microclimatic variation played a stronger role in shaping seedling physiology than species identity. In practical terms, increases in soil moisture and soil temperature were associated with reduced PhiPSII, indicating that temporary waterlogging or warm soils can suppress photochemical performance even in resilient seedlings. Conversely, higher air temperature and humidity were associated with higher PhiPSII, suggesting that seedlings benefited from the warm, humid conditions typical of early summer. Strong location effects in both PhiPSII and gtw further indicate that local habitat conditions, rather than species differences, determine physiological performance, underscoring the importance of microsite selection in regeneration planning. These interaction patterns imply that early seedling success is governed primarily by soil and atmospheric conditions, whereas species differences become evident only under specific combinations of stressors.

The interaction analyses between species and environmental variables (air temperature, air relative humidity, soil moisture content, soil temperature) and VPDleaf revealed clear differences in how PhiPSII (Figure 8) and gtw (Figure 9) varied across environmental gradients for both survivors and non-survivors. To aid visualization, environmental variables and VPDleaf in Figure 8 and Figure 9 are shown as four quartile-based classes (Table S10), summarizing the observed gradients without implying statistical grouping.

Among survivors, PhiPSII increased from the lowest to intermediate air temperature and air humidity classes and declined slightly in the highest classes. Similar patterns were observed for soil moisture and soil temperature, with intermediate levels associated with higher PhiPSII. Differences among species were minor but evident, with Q. pubescens and Q. frainetto maintaining slightly higher PhiPSII under stressful conditions (high VPDleaf and higher soil temperature). At the same time, Q. petraea showed the most substantial difference in PhiPSII from the other three species at high temperature and VPDleaf levels.

Non-survivors showed greater variability and weaker trends across environmental variables. PhiPSII values were lower overall and declined more sharply in the highest temperature, soil temperature, and VPDleaf classes. Q. cerris remained relatively stable across categories, while Q. frainetto showed variable PhiPSII responses. Q. petraea and Q. pubescens exhibited greater sensitivity in the upper environmental classes.

Patterns in gtw are broadly consistent with those in PhiPSII. Among survivors, gtw decreased with increasing temperature and VPDleaf, but increased under more humid and moist conditions, reflecting coordination between gtw and PhiPSII. At moderate soil moisture, the highest gtw value was recorded for Q. pubescens. Non-survivors exhibited lower overall gtw, particularly under hot and dry conditions, suggesting limited capacity to maintain high photosynthetic performance under stress.

Collectively, these results indicate that survivors displayed more consistent and higher physiological performance across environmental classes, whereas non-survivors exhibited steeper reductions in both PhiPSII and gtw. These patterns support H1 and H3 by demonstrating that habitat conditions firmly structured physiological responses and that surviving seedlings maintained more stable physiological performance across environmental gradients.

Mean PhiPSII was strongly and positively associated with survival probability (p < 0.001,

Table 5. Logistic regression results for seedling survival prediction based on mean physiological parameters, species, and location. PhiPS2 was scaled to represent odds ratios per 0.05-unit increase; gtw was converted to mmol m⁻² s⁻¹ to represent odds ratios per 1 mmol m⁻² s⁻¹ increase. (* p < 0.05, *** p < 0.001).

Parameter	Coefficient	SE	Odds Ratio (OR)	95% CI (OR)	p-Value
Intercept	−9.858	1.316	0.000052	0.0000033–0.00059	<0.001 ***
Mean PhiPSII (per 0.05 unit)	0.7056	0.100	2.03	1.68–2.49	<0.001 ***
Mean gtw (per mmol m−2 s−1)	0.0981	0.021	1.1	1.06–1.15	<0.001 ***
Species (ref. Q. cerris)
Q. frainetto	0.5464	0.438	1.73	0.74–4.18	0.212
Q. petraea	−0.2927	0.404	0.75	0.34–1.65	0.469
Q. pubescens	−0.4125	0.416	0.66	0.29–1.50	0.322
Location (ref. Cer)
Debeli Lug	0.9734	0.465	2.65	1.08–6.77	0.036 *
Vraćevšnica	1.9044	0.452	6.72	2.83–16.77	<0.001 ***
Žiča	−0.7081	0.437	0.49	0.21–1.15	0.106

), indicating that seedlings with consistently higher PhiPSII had substantially higher odds of surviving to the end of the study period. Mean gtw showed a similarly strong positive effect (p < 0.001, Table 5), reflecting the importance of maintaining stomatal function under field conditions. Location effects were also significant. Compared with Cer, survival odds were substantially higher at Vraćevšnica and moderately higher at Debeli Lug (Table 5). Žiča showed no statistically significant difference relative to Cer (p = 0.106, Table 5). Species effects were not significant after accounting for PhiPSII, gtw, and location (all p > 0.2), indicating that physiological responses, rather than inherent species differences, were the primary drivers of establishment success.

The model correctly classified 90.8% of seedlings (sensitivity = 97.9%, specificity = 58.8%; Table S11). The high sensitivity indicates that physiological measurements were highly effective at identifying seedlings with high survival probability. In contrast, lower specificity reflects the difficulty of predicting mortality from short-term physiological measurements alone.

4. Discussion

4.1. Habitat Effects on Seedling Physiology (H1)

Previous studies commonly report VPDair, gs/gsw, or Fv/Fm, rather than VPDleaf, gtw, or PhiPSII. Although these metrics are not directly interchangeable, they reflect related aspects of plant water status, stomatal regulation, and photochemical performance and generally respond in similar directions under drought and heat stress. Therefore, studies using these alternative indices remain relevant for interpreting the physiological patterns observed in this study.

Results from this study showed that early post-germination physiological performance of oak seedlings was strongly shaped by environmental differences among locations (fully supporting H1). Physiological parameters (PhiPSII, gtw, VPDleaf) varied more across locations than among species, indicating that microclimate and soil conditions predominate over species differences under the conditions studied. Seedlings that were growing in locations with higher air humidity and lower soil temperatures (e.g., Žiča) had higher PhiPSII (both in the morning and afternoon). In contrast, in warmer/drier locations (e.g., Vraćevšnica), seedlings had reduced physiological performance (in both measurement periods), despite high overall survival. This is consistent with studies showing reduced photosynthetic activity and growth of Quercus seedlings under elevated temperatures and drought [53,54,55]. Similar patterns for stomatal conductance and canopy-level water use have been reported in oaks exposed to high VPD and soil moisture deficit [56,57,58].

The strong location effect was revealed by both LME and ANOVA analyses, indicating that environmental variables were the primary drivers of physiological activity. Soil moisture consistently had a negative effect on PhiPSII, particularly during afternoon measurements, consistent with previous evidence that progressive water stress reduces photosynthetic activity in Mediterranean oak species [59,60]. Comparable declines in physiological performance under drought and rehydration cycles have also been reported for other oak species [61,62]. Air relative humidity and air temperature had significant positive effects on PhiPSII, likely because higher air humidity reduces VPD and facilitates gas exchange in oak seedlings under summer drought [63,64,65]. Similar relationships between atmospheric moisture, stomatal regulation, and drought responses have been documented for oaks and other temperate broadleaves [66,67].

Although elevated VPDleaf and air temperatures reduced physiological performance at Vraćevšnica, seedling survival remained exceptionally high (91.3%). This apparent paradox suggests that short-term reductions in PhiPSII and gtw did not necessarily translate into mortality when other conditions were favorable. Late-season soil moisture remained relatively stable at this site, which may have mitigated hydraulic stress and prevented prolonged xylem tension [61,67,68].

Reduced gtw may reflect a conservative water-use strategy that protects seedlings during episodic stress rather than indicating irreversible damage, a well-documented response in Quercus species [26,62,69]. These patterns suggest that survival depends not only on instantaneous physiological activity but also on the capacity to tolerate transient stress and recover when conditions improve [19,70].

This interpretation is further supported by the logistic regression analysis, which indicated 6.7-fold higher odds of survival at Vraćevšnica compared to Cer. However, within the physiologically monitored subset of seedlings, mortality (12%) was slightly higher than in the whole seeded population (8.7%). This suggests that the measured individuals may have occupied relatively more stressful microsites. Such fine-scale spatial heterogeneity in soil moisture and microclimate is inherent in field plots and likely contributed to the lower physiological performance observed in the monitored subset despite high overall plot-level survival.

Although provenance effects were minimal in our variance partitioning analysis, uneven representation and mortality limited the robustness of inference. The exclusion of provenance from final models represents a design limitation rather than solely a statistical constraint.

4.2. Differences in Physiological Responses Among Oak Species (H2)

H2 was supported, as physiological responses differed among oak species during early post-germination development, although habitat conditions strongly modulated these differences. Results showed that Q. petraea consistently exhibited lower PhiPSII across all locations, reflecting its known sensitivity to drought and elevated evaporative demand [26,71,72]. However, Q. cerris and Q. frainetto maintained the highest photosynthetic activity, even under warmer/drier conditions, consistent with other studies [60,73,74]. Q. pubescens revealed a significant intermediate stress response, performing better than Q. petraea but below Q. cerris and Q. frainetto under harsher conditions. This aligns with existing knowledge of Q. pubescens, which can maintain moderate physiological activity and flexible stomatal control under xeric conditions [75,76,77], suggesting that its resilience is expressed primarily under moderate rather than extreme stress.

Although gtw and VPDleaf showed weaker species differences, the significant interaction between species and location for PhiPSII indicates that species diverged in their physiological responses depending on habitat conditions. These species-level patterns should be interpreted cautiously, as reduced mortality resulted in unequal sample sizes across locations, thereby increasing uncertainty for species represented by fewer individuals. Nevertheless, the consistency of trends across locations and measurement periods suggests that these species-specific differences are robust.

4.3. Morphological Parameters

Morphological variation provided crucial context for interpreting physiological patterns. Q. cerris and Q. frainetto produced taller seedlings than the other two species, while Q. cerris developed the thickest stems, a trait associated with early mechanical stability and stress tolerance [78,79,80]. Location variation mirrored local conditions: seedlings at Cer and Žiča were taller, while those at Debeli Lug and Vraćevšnica invested more in stem diameter. Height-diameter patterns often reflect a trade-off between vertical growth for light capture and radial growth for structural support [81,82,83]. Together with the physiological data, these morphological trends suggest that seedlings tailor their structural development to match the specific demands of each habitat.

4.4. Physiological Predictors of Survival (H3)

Seedlings with higher PhiPSII and more stable stomatal conductance were far more likely to survive the first growing season, strongly supporting H3. Survivors consistently maintained functional physiological performance with PhiPSII values typically exceeding 0.65 and moderate gtw. At the same time, non-survivors exhibited extremely low PhiPSII values (often <0.50, frequently approaching zero), indicating severe photoinhibition, hydraulic failure, or both [27]. Similar declines in PSII electron transport preceding mortality have been reported in drought-exposed Quercus seedlings [67,68,69]. Experimental and field studies further emphasize that summer drought can firmly depress physiological performance and increase mortality risk in oak seedlings [63,84].

The linear mixed-effects models showed that environmental stressors (particularly low air humidity, high air temperature, and reduced soil moisture content) had disproportionately strong negative effects on PhiPSII and gtw in the individuals that ultimately died [26,62]. In contrast, surviving seedlings maintained comparatively stable levels of PhiPSII under the same conditions, highlighting physiological resilience rather than morphological traits as the principal predictor of short-term establishment success [70,85].

PhiPSII has been proposed as an early indicator of stress tolerance and mortality risk in regeneration studies [19,86], and findings from this research support its utility for monitoring direct-seeding outcomes. The logistic regression analysis further reinforces these conclusions, demonstrating that early physiological performance is a strong predictor of survival. The logistic regression revealed that PhiPSII (OR = 2.03 per 0.05 unit, p < 0.001) and gtw (OR = 1.10 per mmol m⁻² s⁻¹, p < 0.001) were strong predictors of survival, with variance inflation factors (VIFs) < 2 confirming no multicollinearity between predictors. Even modest differences in PhiPSII and gtw translate into substantial differences in survival probability. The model’s high classification accuracy (90.8%) and very high sensitivity (97.9%) indicate that physiological parameters reliably identify seedlings with a high probability of survival. The low specificity (58.8%) indicates that some seedlings exhibiting reduced physiological performance nevertheless survived, possibly due to their capacity for recovery. This finding highlights the practical value of physiological monitoring during the early post-germination period as an early-warning tool, enabling adaptive management before mortality occurs. Future research should evaluate whether repeated measurements throughout the growing season improve mortality prediction.

4.5. Practical Implications for Oak Restoration

These findings demonstrate that direct seeding can achieve high first-year establishment, especially under favorable microclimatic conditions, consistent with successful trials in other oak species [11,87,88]. Additional studies have shown that sowing date, acorn traits, and site preparation strongly influence direct-seeding outcomes across different Quercus species [89,90]. However, the patterns observed here should be interpreted within the context of a single growing season and moderate environmental stress. Longer-term studies are needed before concluding performance under sustained drought or under future climate projections. A key implication of this study is that matching species to suitable microsites is more critical than species differences alone. Even highly drought-tolerant species performed poorly when exposed to unfavorable microclimate, while mesic sites supported high establishment success across all species.

Species-specific patterns, while secondary to site effects, still provide guidance for restoration planning. Q. cerris and Q. frainetto showed consistently strong establishment across various environments, suggesting they are suitable candidates for restoration on sites expected to face higher heat or periodic drought. Q. pubescens, although intermediate in survival, displayed stable physiological performance, indicating potential for use in xeric or warming habitats where moderate stress tolerance is sufficient. However, Q. petraea showed reduced establishment on stressful sites, implying that direct seeding of this species will require mesic microsites and shading.

Overall, this study demonstrates that physiological measurements (e.g., PhiPSII, gtw) can provide valuable early-warning indicators of establishment success. Such measurements could be integrated into monitoring protocols to guide adaptive management during the first growing season, improving overall regeneration success under field conditions [88].

5. Conclusions

This study demonstrates that microclimatic and soil conditions strongly determine early physiological performance and survival of directly seeded oak seedlings. Site effects outweighed species effects, highlighting the importance of microsite selection in restoration planning. Aligning species-specific tolerances with favorable microsite conditions is essential, but microsite quality itself was the dominant driver of establishment success. Among species, Q. cerris and Q. frainetto showed the greatest resilience, Q. pubescens performed moderately well, and Q. petraea was the most sensitive to stressful conditions. Seedlings with higher PhiPSII and stable stomatal conductance had noticeably better survival, indicating that early post-germination physiological measurements can serve as reliable indicators of establishment success.