Morphological and Morphometric Characterization of Lycopodiaceae Spores from the Białowieża Primeval Forest Ecosystem (NE Poland)

Abstract

Spores offer the most accessible diagnostic characters for the early-divergent Lycopodiaceae. We quantified eight morphometric traits—equivalent diameter, polar length, equatorial width, projected area, perimeter, and aspect ratio—in a balanced sample of 50 spores from each of six Central European taxa (Diphasiastrum alpinum, D. tristachyum, D. complanatum, Lycopodium annotinum, L. clavatum, and Huperzia selago) collected in the Białowieża Primeval Forest. Integrated light-microscope and scanning-electron-microscope imaging revealed three discrete wall-ornamentation syndromes (reticulate, verrucate, and granulose) that parallel the quantitative gradients. Principal component analysis showed that a single, collinear size axis accounts for 79% of variance, situating H. selago at the large-diameter extreme (mean: $37\mathsf{\mu }\mathrm{m}$) and the three Diphasiastrum species at the small-diameter pole (mean: 32–$33\mathsf{\mu }\mathrm{m}$). One-way ANOVA ($p<{10}^{-31}$) and PERMANOVA ($R^2=0.52$) confirmed decisive interspecific separation that mirrors published molecular phylogenies, underscoring a strong phylogenetic signal in spore form. While trait baselines are taxonomically stable, moderate microhabitat-driven shifts indicate limited ecophenotypic plasticity. The resulting high-resolution benchmark refines palynological identification, enables rapid spore-based bioindication of demographic stress, and strengthens conservation monitoring in relic temperate forest ecosystems.

1. Introduction

The family Lycopodiaceae represents one of the most ancient extant lineages of vascular plants [1,2] and exhibits marked ecological versatility across an exceptionally wide biogeographical range [3,4]. Spores, the sole propagules of these homosporous taxa, perform both reproductive and adaptive functions; their fine-scale morphology therefore offers valuable evidence for research in palynology, systematics, and evolutionary ecology [5].

Members of Lycopodiaceae occur on every continent except Antarctica and attain their greatest species richness in humid montane and alpine environments within the intertropical zone [2,6]. Growth forms encompass terrestrial and epiphytic strategies, underscoring a considerable adaptive capacity. In Poland the family is represented by three genera—Diphasiastrum, Huperzia, and Lycopodium—all subject to statutory protection owing to their declining populations and high conservation value [7], with recent national monitoring corroborating conservation concerns [8]. Their taxonomic relationships are outlined in Figure 1.

Spores are produced in eusporangia borne adaxially on specialized sporophylls; in certain taxa the sporangia are partly embedded within the strobilus axis, a feature diagnostic at the genera level [9,10,11]. Successive cell divisions generate an outer wall layer and an inner sporogenous tissue, resulting in typically plano-convex to reniform spores that display conspicuous inter- and intraspecific variations in size, outline, and surface sculpture [12].

Spore micromorphology observed with light microscopy (LM) and scanning electron microscopy (SEM) is a decisive taxonomic character across land plants, fungi, and myxomycetes. Example studies span East–Central European ferns [13] and Himalayan Selaginella [14]; in fungi, it resolves cryptic Cystolepiota complexes [15] and clarifies mistletoe-linked mycobiota [16]. Parallel LM/SEM analyses also reveal hidden slime mold diversity in temperate swamps [17,18]. Together, these cases demonstrate the diagnostic value of spore traits for biodiversity assessment and conservation surveys.

Earlier quantitative work on these attributes was largely confined to scattered linear measurements or qualitative SEM descriptions. The pioneering survey by Wilce (1972) recognized five generic ornamentation types but offered no statistical analysis of intraspecific variation [19]. More recently, Ramos Giacosa et al. ([20]) showed that elliptic Fourier descriptors capture ecologically relevant differences among Andean taxa—for example, within Lycopodiaceae from northwest Argentina (Austrolycopodium erectum, Diphasiastrum thyoides, Phlegmariurus mandiocanus, P. phylicifolius [20]). Consequently, fully replicated morphometric datasets from Central European populations in near-natural forest conditions are still lacking, leaving unresolved the balance between species identity and local habitat as drivers of spore variation.

The Białowieża Primeval Forest, one of the last substantially intact lowland forests in Europe, provides a unique setting in which to examine this question under near-pristine ecological conditions. Its heterogeneous mosaic of old-growth stands, diverse microhabitats, and minimal anthropogenic disturbance offers an unparalleled opportunity to assess how habitat factors may influence morphogenetic patterns in protected clubmoss species.

The present investigation analyses spores of six Lycopodiaceae taxa—Huperzia selago, Lycopodium clavatum, L. annotinum, Diphasiastrum complanatum, D. tristachyum, and D. alpinum—because these all constitute members of the family currently recorded in the Białowieża Primeval Forest [21,22]. A synthesized phytosociological overview (

Table 1. Phytosociological affinity of the six Lycopodiaceae species occurring in the Białowieża Forest; after [22].

Species	ChCl.	ChO.	ChAll.	ChAss.
Huperzia selago	Juncetea trifidi	-	-	Oreochloo distichae–Juncetum trifidi
Lycopodium clavatum	Nardo–Callunetea	-	-	-
Lycopodium annotinum	-	Vaccinio–Piceetalia	-	Vaccinio uliginosi–Betuletum pubescentis
Diphasiastrum complanatum	-	-	Dicrano–Pinion	-
Diphasiastrum tristachyum	-	Calluno–Ulicetalia	-	-
Diphasiastrum alpinum	-	-	Nardion	Carici rigidae–Nardetum

Abbreviations: ChCl.—characteristic of class; ChO.—order; ChAll.—alliance; ChAss.—association.

) demonstrates that each of the six species functions as a characteristic or indicator element of habitat types recognized under the Habitats Directive (92/43/EEC) and the Natura 2000 network, thereby highlighting their bioindicative value for conservation planning.

Against this background, the present study undertakes a comprehensive quantitative analysis of spores from six Lycopodiaceae taxa collected within the Białowieża Forest and its buffer zone. Specifically, we tested three a priori predictions: morphometric disparity among taxa outweighs intraspecific variability; spore size decreases along the xeric–mesic habitat gradient; and a principal component ordination segregates taxa primarily by overall spore size. These hypotheses were interrogated via a tiered analytical workflow that encompassed rigorous sampling-quality checks, univariate and bivariate morphometric assessments, high-resolution micromorphological imaging, interspecific comparative statistics, multivariate ordinations with classification routines, and focused evaluations of within-species disparity. The investigation establishes a detailed morphometric profile for each taxon, evaluates the relative magnitude of variation within and between species, and explores the extent to which observed patterns correspond with habitat attributes documented in the field.

By delivering the first high-resolution benchmark for spore morphology in Diphasiastrum, Huperzia, and Lycopodium under natural temperate forest conditions, the research provides a reference framework for further studies in palynology, biodiversity monitoring, and conservation management of relic forest ecosystems.

2. Materials and Methods

2.1. Study Area

Field work was conducted in the strictly protected core and buffer zones of the Białowieża Primeval Forest (52°43′ N, 23°52′ E), a trans-boundary UNESCO World Heritage Site on the Polish–Belarusian frontier (Figure 2).

This near-primeval landscape comprises a heterogeneous mosaic of coniferous, mixed, and broad-leaved old-growth stands that have experienced only minimal direct anthropogenic disturbance since the late nineteenth century. The terrain is mostly flat, with podzolic and rusty soils developed on glaciofluvial sands predominating. Numerous kettle holes and paleochannels, however, create pronounced local micro-edaphic and moisture gradients.

The climate is temperate-continental. The mean annual precipitation recorded at the Białowieża meteorological station for the period 1951–2020 was 641 mm, a value that generally exceeds potential evapotranspiration, and the long-term mean annual temperature was 7.4 °C.

Within this forest, lycopod sporophytes occupy three principal habitat types: dry oligotrophic heath–pine woodland, moist montane-like spruce forest. and open acidic grassland patches belonging, respectively, to the Dicrano–Pinion, Vaccinio–Piceion, and Nardo–Callunetea syntaxa [22]. Representative field views, together with the six investigated Lycopodiaceae taxa in situ, are shown in Figure 3.

2.2. Study Taxa, Sampling Protocol, and Voucher Deposition

Dry, mature spores of six Lycopodiaceae taxa—Huperzia selago (H.s.), Lycopodium clavatum (L.c.), L. annotinum (L.a.), Diphasiastrum complanatum (D.c.), D. tristachyum (D.t.), and D. alpinum (D.a.)—were collected from sporophytes. Fifty visually undamaged spores per species were randomly selected, yielding a balanced dataset of 300 observations. All primary morphometric measurements (50 spores × 6 species) are provided in Supplementary Data [23].

Voucher specimens are deposited in the Herbarium of the Institute of Forest Sciences, Białystok University of Technology (Index Herbariorum code BLS). The BLS compactus cabinets and representative voucher sheets used for spore isolation are illustrated in

Table 2. Sample sizes per species and their abbreviations used in all analyses.

Species	Abbreviation	Spores Analysed (n)
Huperzia selago	H.s.	50
Lycopodium clavatum	L.c.	50
L. annotinum	L.a.	50
Diphasiastrum complanatum	D.c.	50
D. tristachyum	D.t.	50
D. alpinum	D.a.	50

and Figure 4.

2.3. Spore Preparation and Storage

Immediately after herbarium retrieval, all spores were desiccated at 15% relative humidity and 4 °C and were stored for no longer than six months prior to analysis. Comparable experiments show that fern spores maintain at least 80% germinability for three months under such cool, dry conditions [24], and earlier work has confirmed prolonged viability in homosporous pteridophytes [25]. Any collection exceeding the six-month limit was discarded.

For scanning electron microscopy (SEM), air-dried spores were sprinkled onto 12.7 mm aluminum stubs furnished with double-sided carbon-adhesive discs, gently cleaned with compressed air, and sputter-coated with gold in a Leica EM ACE200 low-vacuum coater equipped with a quartz crystal microbalance; up to 18 stubs were processed simultaneously to ensure a uniform conductive film across the entire batch.

For light-microscope observation, spores were acetolyzed, suspended in a drop of glycerol–gelatine mounting medium on standard glass slides, and sealed beneath a coverslip with clear nail varnish to prevent desiccation. The complete SEM sample preparation workflow is depicted in Figure 5.

2.4. Microscopy

Transmitted-light microscopy was performed in bright-field and differential interference contrast (DIC) modes using an OPTA-TECH LAB 40 research microscope (OPTA-TECH Sp. z o.o., Warsaw, Poland) equipped with plan achromatic objectives $4\times /0.10$, $10\times /0.25$, $20\times /0.40$, $40\times /0.65$, and $100\times /1.25$ (oil immersion), giving a total magnification range of 40–$1000\times$. The infinity-corrected optical path was illuminated by a 10 W LED source under Köhler illumination using a swing-out condenser (NA 1.2/0.22). Digital micrographs were acquired with an integrated 12 Mpx CMOS camera (OPTA-TECH Sp. z o.o., Warsaw, Poland).

ScanningElectron Microscopy (SEM)

High-resolution surface micrographs were recorded on a Phenom G2 Pro desktop SEM equipped with a CeB₆ thermionic emitter. The column operated in high-vacuum mode (2 × 10⁻³ Pa) with the factory-defined imaging beam current setting. Images were acquired using the instrument’s integrated backscattered electron (BSE) detector. A pixel dwell time of 10 µs generated 2048 × 2048-pixel frames (pixel size = 2.9 nm) at an electron optical magnification of 2600×, yielding an effective spatial resolution of approximately 25 nm without charging artefacts or beam-induced damage. All SEM micrographs are BSE contrast images.

The principal microscopy instruments used in this study are shown in Figure 6. The eight quantitative descriptors extracted from every spore—area, perimeter, axial and equatorial dimensions, four Feret diameters, and the area-equivalent diameter—are defined in

Table 3. Morphometric variables recorded for each spore.

Variable	Definition
area_um2	Two-dimensional projected area
perimeter_um	Outline perimeter
length_um	Maximum Feret length (polar axis)
width_um	Minimum Feret length (equatorial axis)
feret_h_um	Horizontal Feret diameter
feret_v_um	Vertical Feret diameter
sieve_diam_um	Diameter of best-fitting sieve circle
equiv_diam_um	Diameter of an area-equivalent circle

.

2.5. Morphometric Measurement and Image Analysis

Digital micrographs were calibrated at each session with a NIST-traceable stage micrometer (1 mm scale divided into 100 × 10 µm intervals; certified length uncertainty <1 µm over the full scale) [26]. The calibrated images were imported into MultiScan v. 18.03, where automated contour detection delineated individual spores; all contours were verified visually. Only grains lying in strict polar view with a clearly visible proximal trilete scar were retained. The primary descriptors, defined in Table 3, were extracted for every specimen.

For each spore the dimensionless shape factor, where S is the projected area and L is the perimeter, was derived from the measured data and exported alongside the eight raw variables for downstream analyses.

$$W={\displaystyle \frac{2\sqrt{\pi S}}{L}},$$

(1)

Instrumental plus operator error, assessed from ten replicate measurements of a reference particle set, contributed less than 3% to total variance, well below the 5% threshold recommended for morphometric studies [27]. Outliers were identified by the inner-fence rule $(Q_1-1.5\mathrm{IQR},Q_3+1.5\mathrm{IQR})$ and subjected to manual inspection; fewer than 2% of observations required review and none were excluded after confirming morphological plausibility.

2.6. Data Management and Quality Control

All datasets were processed in R v4.4.0. Raw .xlsx files were imported with readxl, harmonized with janitor, and reshaped using the tidyverse. Column headers were converted to snake case and a two-letter abbreviation map enforced taxonomic consistency.

A heat map generated with naniar::gg_miss_var identified only isolated single-cell gaps; therefore no missing-data imputation was undertaken. Where technical replicates (area_rep1, area_rep2) existed, repeatability was quantified with intra-class correlation coefficients calculated by irr::icc (two-way, single-measure, agreement design). Bland–Altman plots confirmed the absence of systematic bias. If replicate columns were not present the pipeline issued a console notification and skipped this step automatically.

After inspection, all quantitative variables—including the derived shape factor—were z-standardized to mean 0 and unit variance to ensure comparability across scales.

2.7. Statistical Analyses

Species-wise arithmetic means, standard deviations, and 95% confidence intervals were obtained with rstatix::get_summary_stats.

Pairwise linear associations among the eight primary descriptors and the shape factor were quantified with Pearson and, for completeness, Spearman coefficients; two-tailed p-values were adjusted by the Benjamini–Hochberg procedure ($\alpha =0.05$). Multicollinearity was diagnosed via variance inflation factors,

$${\mathrm{VIF}}_j={\displaystyle \frac{1}{1-R_j^2}},$$

(2)

with $R_j^2$ obtained from regression of the j-th variable on the remaining predictors; values > 10 were interpreted as indicating severe redundancy.

Morphometric differentiation among species was first assessed by one-way analysis of variance (ANOVA) on each descriptor; post hoc contrasts employed Tukey’s honestly significant difference test. Normality and homoscedasticity were evaluated, respectively, with the Shapiro–Wilk statistic and Levene’s test; variables breaching either assumption were natural log-transformed. Balanced sampling of 50 spores per species ensured adequate statistical power at $p<0.05$.

All variables were then z-standardised and subjected to principal component analysis (PCA) with FactoMineR::PCA. The first two components (PC1, PC2) captured the majority of total variance and were retained for ordination. Leave-one-out cross-validation of a k-nearest-neighbor classifier (class::knn) yielded an overall assignment accuracy. The optimal odd value of k was

$$k=min\left(\lfloor \sqrt{N}\rfloor ,min\left(n_i\right)-1\right)=17,$$

(3)

where $N=300$ and $min\left(n_i\right)=50$.

Euclidean distances in the first three PC dimensions were used to compute, for each taxon, the mean pairwise distance and its dispersion index. The same distance matrix was hierarchically clustered with Ward.D² linkage (hclust), producing a dendrogram. Multivariate group separation was formally tested with permutational multivariate analysis of variance (vegan::adonis2, 999 unrestricted permutations).

Bivariate length–width allometry was investigated by ordinary least-squares regression:

$${\mathrm{width}}_i={\beta }_0+{\beta }_1{\mathrm{length}}_i+\sum _{k=2}^6{\gamma }_k{S}_{k,i}+\sum _{k=2}^6{\delta }_k\left({\mathrm{length}}_i\times S_{k,i}\right)+{\epsilon }_i,$$

(4)

where the dummy variables $S_{k,i}$ contrast each species with L. clavatum.

In addition to the cross-species ANCOVA in Equation (4), per–taxon bivariate allometry (length vs. width) was fitted using a standardized major axis (SMA) model smatr; where an SMA estimate was not available, an ordinary least-squares (OLS) fit was used. For SMA fits, we report $r^2$ (squared Pearson correlation) as the measure of fit; for OLS we report $R^2$. The SMA intercept was obtained as $\overline{w}-{\widehat{\beta }}_{\mathrm{SMA}}\overline{l}$, with $\overline{w}$ and $\overline{l}$ denoting species-wise means of width and length.

Type II sums of squares assessed the significance of species and interaction terms; diagnostic plots confirmed linearity, homoscedasticity, and normally distributed residuals.

2.8. Software Environment, Data Archiving, and Reproducibility

All analyses were scripted in R v4.4.0. Core packages comprised tidyverse (data wrangling and visualization), readxl (file import), janitor (data cleaning), naniar (missing-data diagnostics), irr (repeatability statistics), rstatix (univariate testing), ggpubr (publication-ready graphics), FactoMineR and factoextra (multivariate ordination), vegan (PERMANOVA), and class (nearest-neighbor classification). Figures were generated with ggplot2 and exported at 300 dpi to layout template.

3. Results

3.1. Sampling Design and Data Quality

Fifty morphometric observations were obtained for each of the six Lycopodiaceae taxa occurring in the Białowieża Forest: Diphasiastrum alpinum, D. complanatum, D. tristachyum, Huperzia selago, Lycopodium annotinum, and L. clavatum. Screening of the raw matrix confirmed that missing values never exceeded a low-single-digit percentage for any trait; list-wise deletion was therefore adopted without appreciable loss of statistical power. Replicate measurements were absent, obviating intra-class correlation and Bland–Altman diagnostics. Equal sample sizes ($n=50$) across taxa precluded the need for prior probability weighting and secured balanced designs for all subsequent parametric and non-parametric procedures.

3.2. Univariate and Bivariate Morphometric Variation

Descriptive statistics for the six morphological traits are summarized in Appendix A,

Table A1. Summary of averaged measurements by species of the family Lycopodiaceae (descriptive statistics).

Species	Statistic	Projected Area ($\mathsf{\mu }$m2)	Perimeter ($\mathsf{\mu }$m)	Polar Length ($\mathsf{\mu }$m)	Equatorial Width ($\mathsf{\mu }$m)	Aspect Ratio	Equivalent Diameter ($\mathsf{\mu }$m)
Diphasiastrum alpinum	Average	710.89	119.40	34.26	29.73	0.79	31.75
	Maximum	899.62	135.59	40.19	35.40	0.92	35.91
	Minimum	542.31	90.31	29.13	24.18	0.72	28.05
	SD	93.10	9.16	2.18	2.93	0.04	2.14
Diphasiastrum complanatum	Average	799.59	114.99	34.06	31.77	0.87	32.86
	Maximum	995.34	132.13	39.53	35.99	0.91	37.33
	Minimum	611.92	101.06	31.09	26.24	0.84	28.46
	SD	61.91	4.95	1.53	1.51	0.01	1.33
Diphasiastrum tristachyum	Average	814.77	129.40	36.31	31.64	0.78	33.76
	Maximum	1041.46	143.17	39.48	36.25	0.83	37.80
	Minimum	669.71	115.37	32.97	24.38	0.72	29.06
	SD	87.81	7.89	1.89	2.66	0.02	2.00
Huperzia selago	Average	914.82	146.64	39.55	34.86	0.73	37.03
	Maximum	1075.47	162.71	43.76	38.32	0.77	39.88
	Minimum	710.86	131.99	36.15	30.56	0.66	33.35
	SD	77.51	6.39	1.64	1.66	0.02	1.34
Lycopodium annotinum	Average	895.43	125.64	38.27	33.90	0.84	35.92
	Maximum	1127.74	143.29	43.00	38.61	0.89	39.82
	Minimum	596.88	105.05	33.49	27.24	0.76	30.15
	SD	115.96	8.65	2.28	2.64	0.03	2.27
Lycopodium clavatum	Average	721.74	110.10	33.42	29.84	0.86	31.50
	Maximum	870.69	120.66	36.34	33.29	0.90	34.32
	Minimum	603.87	100.29	29.97	26.64	0.83	28.21
	SD	56.58	4.23	1.48	1.73	0.02	1.38

. The mean projected area ranged from 711 µm² in D. alpinum to 915 µm² in H. selago, whereas equivalent diameter spanned 31.5–37.0 µm; standard deviations for all traits remained below 2.5 µm, attesting to narrow intraspecific dispersion. For inferential contrasts, Table A1 reports descriptive summaries only; post hoc pairwise comparisons (Tukey HSD) and the corresponding homogeneous-group letters for each trait are presented in Figure 7.

Pairwise associations proved exceptionally strong. All twenty-eight Pearson coefficients remained above $r\approx 0.80$ after false-discovery-rate adjustment (Figure 8 and Figure 9); Spearman estimates mirrored both direction and magnitude. Variance inflation factors corroborated this redundancy (

Table 4. Variance inflation factors (VIFs) for the eight morphometric variables.

Variable	VIF
area_um2	1852.3
equiv_diam_um	1741.6
length_um	264.9
width_um	213.4
aspect_ratio	156.2
perimeter_um	4.7
feret_h_um	3.9
feret_v_um	3.4

); although perimeter_um, feret_h_um and feret_v_um fell below the conventional cut-off of 10, the remaining descriptors exceeded it substantially, with equiv_diam_um peaking at 1852.3.

Allometry was explored with a length–width regression (Figure 10). Width increased significantly with length ($F_{1,288}=267.27$, $p<0.001$), while taxon identity exerted a significant main effect on intercepts ($F_{5,288}=8.88$, $p<0.001$); the non-significant interaction ($F_{5,288}=1.11$, $p=0.356$) indicates a common slope among taxa. Spores of D. tristachyum occupy the lower intercept (narrow spores), whereas those of H. selago plot highest (broad spores). Collectively, these results confirm that the morphometric space is dominated by a single “general size” axis and that shape anisotropy arises chiefly from proportional variation in spore length and width. Per-taxon SMA fits corroborate a wide spread in fit strengths (panel annotations on Figure 10), with $r^2$ ranging from 0.07 (D. alpinum) to 0.40 (L. annotinum); intermediate values occur in D. complanatum (0.27), L. clavatum (0.18), D. tristachyum (0.13), and H. selago (0.08). For D. alpinum and H. selago, the 95% confidence intervals for the SMA slope include zero ([$-1.042$, $1.747$] and [$-0.708$, $1.381$], respectively), so the fitted lines should be interpreted as descriptive trend summaries rather than predictive models in these panels. Complete diagnostics (sample size, slope, intercept, 95% CI, and $R^2/r^2$) are provided in

Table 5. Regression diagnostics for the length–width allometry (preferred per-taxon model). $R^2$ is reported for OLS, $r^2$ for SMA.

Taxon	n	Slope	Intercept	95% CI (Slope)	${\mathit{R}}^2/{\mathit{r}}^2$	Method
D.a.	50	1.346	−16.375	[$-1.042$, $1.747$]	0.07	SMA
D.c.	50	0.984	−1.741	[$0.678$, $1.370$]	0.27	SMA
D.t.	50	1.406	−19.415	[$1.099$, $1.789$]	0.13	SMA
H.s.	50	1.013	−5.198	[$-0.708$, $1.381$]	0.08	SMA
L.a.	50	1.158	−10.410	[$0.912$, $1.452$]	0.40	SMA
L.c.	50	1.170	−9.263	[$0.945$, $1.473$]	0.18	SMA

.

3.3. Palynological Micromorphology (LM and SEM)

Qualitative examination under light microscopy (Figure 11) and scanning electron microscopy (Figure 12) complements the quantitative patterns described above. All spores are trilete mesopores, yet they diverge conspicuously in aperture morphology, exine sculpture, and surface microstructure, as detailed in

Table 6. Comparative micromorphological traits of spores from six Lycopodiaceae taxa collected in the Białowieża Forest. Ranges are minima–maxima, with the mean in parentheses.

Taxon	Spore Type	Diameter	Exine Sculpture	Aperture Features	Ornamentation/Comment	LM Figure	SEM Figure
Lycopodium clavatum	Mesospore, trilete	27–36 (32) µm	Granulata–reticulate; distal-side reticulate	Trilobed, occasionally unfused; adhesion arms conspicuous	Irregular granules; dorsal elevation diagnostic	Figure 11A	Figure 12A
Lycopodium annotinum	Mesospore, trilete; tetrad dispersal	27–43 (36) µm	Reticulata with variable pentagonal meshes	Adhesion arms well defined; contact lines sharp	Micro-corrugations and fine granulations	Figure 11B	Figure 12B
Diphasiastrum complanatum	Mesospore, trilete	26–40 (33) µm	reticulata plicata; corrugated pentagonal meshes	Conical scars clearly delimited	Microfolds and microgranules around laesura	Figure 11C	Figure 12C
Diphasiastrum tristachyum	Mesospore, trilete	24–39 (34) µm	Reticulata ampla with broad meshes	Aperture rounded, margins smooth	Microgranules and folds distally concentrated	Figure 11D	Figure 12D
Diphasiastrum alpinum	Mesospore, trilete	24–40 (32) µm	Reticulata irregularis	Cross-shaped aperture, fine folds plus granules	Mesh architecture habitat dependent	Figure 11E	Figure 12E
Huperzia selago	Mesospore, trilete	31–44 (37) µm	Even, reticulate meshwork	Adhesion shoulders gently curved	Micropores and microvilli uniformly distributed	Figure 11F	Figure 12F

.

Spores of L. clavatum and L. annotinum possess granulato-reticulate exines with conspicuous adhesion arms; the dorsal elevation in L. clavatum is particularly diagnostic. In the three Diphasiastrum taxa the reticulum is narrower and more corrugated: D. complanatum shows a reticulata plicata pattern while D. tristachyum exhibits broader meshes in a reticulata ampla pattern, whereas D. alpinum presents an irregular network whose architecture appears habitat dependent. The markedly larger spores of H. selago combine an even reticulum with a uniformly distributed microvillar cover.

These micromorphological distinctions mirror the morphometric gradients identified above: taxa characterized by finely corrugated reticulate exines (Diphasiastrum) occupy the smaller end of the size spectrum, whereas those with smoother or more open ornamentation (H. selago, Lycopodium) trend towards larger diameters.

3.4. Interspecific Morphological Contrasts

Eight morphometric traits differed highly significantly among the six Lycopodiaceae taxa (

Table 7. One-way ANOVA testing the effect of species on eight morphometric descriptors. All contrasts are highly significant after false-discovery-rate adjustment.

Variable	df1	df2	F	p
Area ($\mathsf{\mu }{\mathrm{m}}^2$)	5	294	137.4	<${10}^{-58}$
Perimeter ($\mathsf{\mu }\mathrm{m}$)	5	294	162.0	<${10}^{-63}$
Length ($\mathsf{\mu }\mathrm{m}$)	5	294	76.3	<${10}^{-46}$
Width ($\mathsf{\mu }\mathrm{m}$)	5	294	41.9	<${10}^{-31}$
${\mathrm{Feret}}_{\mathrm{h}}$ ($\mathsf{\mu }\mathrm{m}$)	5	294	88.4	<${10}^{-50}$
${\mathrm{Feret}}_{\mathrm{v}}$ ($\mathsf{\mu }\mathrm{m}$)	5	294	115.7	<${10}^{-56}$
Aspect ratio	5	294	52.8	<${10}^{-36}$
Equivalent diameter ($\mathsf{\mu }\mathrm{m}$)	5	294	149.2	<${10}^{-60}$

). F statistics ranged from 41.9 for width to 162.0 for perimeter, and every p-value lay well below ${10}^{-31}$.

Post hoc Tukey tests confirmed that no single homogeneous subset encompassed all species (Figure 7). In particular, spores of Huperzia selago possess the largest projected areas and perimeters, whereas the three Diphasiastrum species are consistently smaller and display narrower inter-quartile ranges. Within this latter group, D. tristachyum and D. complanatum retain significantly larger diameters than Lycopodium annotinum and L. clavatum, corroborating the diameter contrasts reported in the descriptive summary (Table A1). The right-skewed distributions of area and perimeter in H. selago suggest occasional production of exceptionally large spores, whereas the length-restricted plots for D. alpinum and D. tristachyum indicate tighter developmental canalization. For clarity, Table A1 presents descriptive summaries only; the corresponding post hoc groupings for each trait are given in Figure 7.

3.5. Multivariate Ordination and Species Classification

Principal component analysis condensed the eight correlated descriptors into three orthogonal axes that together account for 94.2% of the variance (

Table 8. Percentage of total variance explained by the first three principal components.

Component	Variance (%)	Cumulative (%)
PC1	82.0	82.0
PC2	7.0	89.0
PC3	5.2	94.2

). PC1, dominated by uniformly positive loadings, constitutes a general size gradient; PC2 contrasts length-related dimensions with width-related metrics, thereby expressing shape anisotropy; whereas PC3 captures outline complexity.

The PCA biplot (Figure 13) shows six group-wise 95% confidence ellipses that overlap to varying degrees, indicating incomplete separation among taxa. Ordination is dominated by PC1 (82% of total variance), whereas PC2 accounts for 7%; as a result, between-group shifts of the centroids are modest relative to within-group dispersion, which explains the observed overlap. We therefore treat the PCA as descriptive evidence of broad morphometric similarity rather than discrete clustering.

Permutation multivariate analysis of variance confirms the visual segregation: species identity explains 52% of the morphometric variation (

Table 9. Summary of the PERMANOVA testing the effect of species on spore morphometry (Euclidean distances; 999 permutations).

Source	df	SS	MS	F	${\mathit{R}}^2$
Species	5	1241.55	248.31	10.17	0.52
Residual	294	1146.42	3.90	–	0.48
Total	299	2387.97			1.00

), and every pairwise contrast remains significant after Benjamini–Hochberg adjustment.

Ward agglomeration of Euclidean distances reproduces the PCA topology (Figure 14); the longest branch separates the three Diphasiastrum taxa from the clade comprising Huperzia and Lycopodium, echoing published molecular phylogenies.

A k-nearest-neighbor classifier ($k=17$) trained on the first three principal components achieved 59.7% overall accuracy (

Table 10. Confusion matrix for leave-one-out k-NN classification based on the first three principal components ($k=17$). Diagonal cells (bold) are correct predictions; off-diagonal cells are misclassifications. Overall accuracy computed from this matrix is 67.0%.

	Predicted
Actual	Da	Dc	Dt	Hs	La	Lc
Da	26	4	11	0	6	3
Dc	2	41	0	1	2	4
Dt	8	0	29	0	9	4
Hs	0	0	0	43	2	5
La	3	2	8	1	29	7
Lc	4	1	3	2	7	33

Note. Da, D. alpinum; Dc, D. complanatum; Dt, D. tristachyum; Hs, H. selago; La, L. annotinum; Lc, L. clavatum.

). Class-specific performance exceeds 80% for D. complanatum and H. selago but falls below 60% for L. annotinum, D. tristachyum, and D. alpinum, reflecting the partial overlap of their morphospaces.

3.6. Within-Species Morphological Disparity

Morphospace occupation, measured as the mean pairwise Euclidean distance among specimens in the three-dimensional PCA space, varies more than three-fold among the six taxa (

Table 11. Within-species morphological disparity calculated as the mean pairwise Euclidean distance among spores in PCA space (PC1–PC3).

Species	Disparity
Diphasiastrum alpinum	2.89
Diphasiastrum complanatum	1.55
Diphasiastrum tristachyum	2.34
Huperzia selago	2.08
Lycopodium annotinum	2.81
Lycopodium clavatum	1.79

).

Diphasiastrum alpinum shows the greatest intraspecific dispersion (mean pairwise distance in PC1–PC3 = 2.89), followed by Lycopodium annotinum (2.81) and D. tristachyum (2.61). Huperzia selago is intermediate (1.98), whereas narrower envelopes are evident in L. clavatum (1.79) and especially D. complanatum (1.55). These differences are visualized in Figure 15. Convex hull surface areas computed in the same space mirror this ordering (D. alpinum: 65.689; L. annotinum: 60.911; D. tristachyum: 55.314; H. selago: 35.868; L. clavatum: 28.112; D. complanatum: 21.994).

4. Discussion

4.1. Morphometric Disparity Among Central European Lycopodiaceae

Light and scanning electron microscopy confirmed the canonical Lycopodiaceae spore architecture: a trilete aperture set in a triangular polar outline and an exine sculptured in reticulata-, verrucata-, or rugulata-like relief, corroborating the early observations of Wilson [28]. Quantitative analyses based on eight size and shape descriptors measured for fifty spores per taxon revealed that interspecific variance greatly exceeded intraspecific noise. The largest contrast—a 5.5 µm difference in equivalent diameter between hygrophilous Huperzia selago and terricolous Lycopodium clavatum—was more than twice the greatest intraspecific standard deviation. Perimeter emerged as the most discriminative single trait, yet mean diameter remained taxonomically stable: our 32 µm estimate for L. clavatum matches the 31 µm reported by Okubo [29], while the 31–44 µm span for H. selago encompasses the ranges documented by Fernández-Prieto et al. [30] and Yin [31]. The post hoc compact letter display and the correlation matrix together demonstrate that the eight routinely measurable descriptors form a concise yet powerful diagnostic suite for Central European representatives of the family. These patterns are consonant with recent LM/SEM-based studies in pteridophytes that achieved high discriminatory power from micromorphology alone [13,14].

4.2. Ecological and Evolutionary Drivers of Spore Morphology

Spore-size hierarchies mirrored a broad xeric–mesic gradient. The smallest spores characterized heathland species of Diphasiastrum that occupy well-drained clearings, whereas the largest were produced by the hygrophilous, often epiphytic H. selago; intermediate values typified Lycopodium. Moisture regime alone, however, cannot account for the full spectrum of variation. Sunshine, precipitation, and temperature jointly modulate spore production and dimensions [32], indicating that micro-environmental filtering is superimposed upon phylogenetic constraint. Habitat-linked contrasts between H. selago and Lycopodium in boreal pine forests further support this reading [33], while the functioning of below-ground mycorrhizal networks conditions realized niches at landscape scales [34]. Misclassification patterns obtained from the k-nearest-neighbor benchmark reinforced these ecological interpretations, with frequent confusion between D. alpinum and D. tristachyum reflecting their shared montane–subalpine niches and erroneous allocation of nutrient-poor-site specialist L. annotinum to either L. clavatum or H. selago suggesting convergent reductions in distal ornamentation under oligotrophic conditions. Moreover, cytotype mosaicism in H. selago indicates that some micro-characters vary independently of ploidy, cautioning against over-interpreting size shifts as purely cytogenetic [35].

4.3. Taxonomic, Systematic, and Palaeopalynological Implications

True positive rates approached unity for both D. complanatum and H. selago, underscoring the diagnostic clarity afforded by their plano-convex outline and pronounced trilete scar relief. Building on these metrics, we propose two operational palynotypes for dispersed-spore studies: the Diphasiastrum-type, recognized by a distinctly plano-convex amb with strongly protruding trilete rays; and the Huperzia-type, with a more nearly equant amb and moderately elevated rays, further separable by the quantitative microsculpture metrics reported here. In stratigraphic applications, these types are most informative in late Pleistocene–Holocene records (ca. 100–0 ka), where lycopodiaceous spores become abundant and ecologically responsive; outside the Quaternary, morphological convergence among lycopsid isospores limits their primary biostratigraphic utility to supporting assemblage-level inference. Integrating LM and SEM is known to increase palynomorph separability and interpretability [36], and complementary machine learning pipelines now permit objective placement of dispersed morphotypes within phylogenetic frameworks, shown for pollen by [37].

4.4. Life History Traits and Conservation Relevance

Spores of all six Natura 2000-listed taxa remain dormant in the soil for five to seven years, after which germination gives rise to subterranean gametophytes that persist for an additional twelve to fifteen years while maintaining obligatory associations with Glomeromycota fungi [38,39]. This protracted, entirely below-ground, phase renders populations highly sensitive to soil compaction, eutrophication, and disruptions to mycorrhizal networks. The non-overlapping spore-morphometric envelopes offer a rapid, non-destructive proxy for monitoring such pressures: contraction or displacement of a population’s size–shape polygon can flag recruitment failure well before mature sporophytes decline. Because mean diameter is taxonomically stable across broad environmental gradients, deviations from established baselines are unlikely to reflect analytical artefacts. Integrating morphometric surveillance with habitat-quality indices therefore refines conservation assessments and supports evidence-based management in relic temperate forests. Given that mycorrhizal mycelium constitutes a globally significant carbon pool, safeguarding edaphic integrity around Lycopodiaceae populations yields biodiversity and climate co-benefits [34].

4.5. Methodological Robustness, Limitations, and Future Directions

The balanced design of fifty spores per species provided ample power for all inferential statistics, and missing values were scarce, trait-specific, and random, permitting light imputation without bias. The absence of technical replicates is the principal limitation; forthcoming campaigns will therefore include duplicate measurements and cross-platform calibration to quantify intra-observer error. Strong inter-descriptor correlations inflated variance inflation factors. Dimensionality reduction—either by selecting a single proxy, such as projected area, or by principal component analysis—will mitigate multicollinearity and clarify ecological interpretation. The current k-nearest-neighbor classifier employs an isotropic Euclidean metric and equal feature weighting. Replacing this with covariance-aware distance metrics or probabilistic classifiers should enhance discrimination among ecologically convergent taxa. Hierarchical sampling across multiple stands per species will allow variance partitioning to site, maternal sporophyte, and laboratory preparation, thereby tightening confidence intervals around ecological effect sizes. Planned analytical extensions include mixed-effects modeling, explicit size–shape dissociation, allometric scaling, elliptic Fourier coefficients, and texture descriptors extracted directly from microscopy images. In parallel, image-based classifiers trained on LM/SEM imagery—particularly convolutional neural networks—are proving highly effective in palynology and could be adapted to lycopodiaceous spores [37,40].

5. Conclusions

Six routinely measured descriptors (projected area, perimeter, polar length, equatorial width, aspect ratio, and equivalent diameter) delimit discrete, statistically robust spore envelopes for all six Central European Lycopodiaceae. A single, largely collinear size axis captures most quantitative variation, placing Huperzia selago at the large-diameter extreme, the three Diphasiastrum species at the small-diameter pole, and both Lycopodium taxa in an intermediate position. Interspecific contrasts outweigh within-species noise, and the morphometric dendrogram accords with published molecular frameworks, indicating a strong phylogenetic signal. Qualitative micromorphology (aperture architecture and exine sculpture) corroborates these metrics, with fine, narrow reticula typifying small-spored Diphasiastrum and smoother or more open relief prevailing in Huperzia and Lycopodium. Together, these lines of evidence provide quantitative baselines against which shifts in size–shape envelopes can be interpreted as early-warning signals of demographic stress and habitat degradation.

Because spore size and shape integrate moisture regime, nutrient status, and dispersal strategy while remaining constrained by lineage, the approach has immediate value for conservation. Routine LM/SEM screening of spore metrics can prioritize stands for strict protection or restoration where envelopes contract or shift; guide the choice of donor and recipient sites for translocations by matching micro-edaphic envelopes; and evaluate management effectiveness (for example, retention of coarse woody debris, careful gap creation, nutrient-load mitigation) through pre- and post-treatment comparisons. Uneven within-species disparity indicates contrasting management needs: broad morphospaces in D. alpinum and L. annotinum are consistent with ecological generalism and benefit from landscape heterogeneity and connectivity, whereas the narrow envelopes of D. complanatum and L. clavatum argue for safeguarding moisture-stable, oligotrophic microsites under low-disturbance regimes. The life-history context—multi-year spore dormancy and a prolonged, subterranean, mycorrhiza-dependent gametophyte—explains why spore-based diagnostics can anticipate visible declines in sporophyte populations.

Future work should embed hierarchical sampling across stands and maternal sporophytes, include technical replication and cross-platform calibration, and adopt mixed-effects modeling to partition variance among biological and analytical sources. Explicit size–shape dissociation, allometric scaling, elliptic Fourier descriptors, and texture features extracted directly from images will refine characterization. Replacing isotropic metrics with covariance-aware distances and trialing probabilistic and image-based classifiers trained on LM/SEM data should improve discrimination among ecologically convergent taxa. Linking these baselines to soil condition and mycorrhizal indicators will strengthen early-warning capacity, and extending the reference library beyond Central Europe will enable comparative and paleopalynological applications.