Are morphological characteristics of Parrotia (Hamamelidaceae) pollen species diagnostic? ☆

preferences


Introduction
Some areas of the northern Hemisphere were less impacted by the severe cold climate conditions of the Pleistocene glaciations (Frenzel et al., 1992).For that reason, particular areas in the temperate biome of Eurasia became refugia for several plant species referred to as Arcto-Tertiary floral elements (Manchester et al., 2009).Species such as Zelkova carpinifolia (Pall.)Dippel, Parrotia persica (DC.)C.A.Mey., Pterocarya fraxinifolia (Lam.)Spach, Acer cappadocicum Gled., and Ginkgo biloba L. are valid examples of relict plant taxa that nowadays occur in refugia regions in Iran (the Hyrcanian forest), on Crete (Greece) and Sicily (Italy), and in Eastern Asia.However, while these plants are important in characterizing the impact of the last glacial/interglacial cycles (Manchester et al., 2009;Akhani et al., 2010), the paleoclimatic and paleophytogeographic history of these relict species in their modern refugia is poorly understood (Cao et al., 2016).The lack of information is partly attributed to the scarcity of Cenozoic plant fossil records from the Caucasus prior to the Late Miocene (Shatilova et al., 2011).In addition, the difficulty in identifying tricolpate pollen of Hamamelidaceae is considered the cause of the absence of Parrotia in Eastern-Asia Cenozoic floras, a region which is now home to many Arcto-Tertiary floral elements (Zhi-chen et al., 2004).Moreover, due to the dampened effect of Quaternary glaciations the Arcto-Tertiary elements persisted in their ecosystems and only developed slight morphological changes which might mask recent diversification events (Qian and Ricklefs, 2000;Nagalingum et al., 2011).Thus, a focus on the detection and characterization of morphological differences and variability between species of the same genus, from the past to present, will provide insights into the evolutionary history of these Arcto-Tertiary floral elements in response to for example, the formation of dispersal barriers and environmental changes.
Pollen is primordial for highlighting changes in plant taxa from both a chronological and chorological perspective (Marquer et al., 2014;Martin and Harvey, 2017).Morphological dissimilarities between pollen of closely related species may also reflect differences in the adaptation to long-term environmental changes (e.g.Cruden, 1976;Hedhly et al., 2005).Several paleoecological studies, especially from Europe, provide insight into Quaternary environments, vegetation, and climate (e.g., Leroy and Roiron, 1996;Reille et al., 2000), and some studies have focused on specific relict plant taxa, such as Zelkova (Follieri et al., 1986) and Parrotia (Bińka et al., 2003), from particular geographic regions.However, the limitation of the palynological approach is that pollen identification can usually be made only at the family and/or genus level and it can be hard or impossible to differentiate species or subspecies of the same genus (Nakagawa et al., 1998, for Zelkova spp.).For example, the pollen of Zelkova has been discovered from Quaternary sediments in most of Europe, and these pollen records could represent several different species, which some have gone extinct and/or are now relicts in Crete (Kozlowski et al., 2014) and Sicily (Garfì and Buord, 2012).The lack of ability to distinguish pollen from different sister species prevents reliable biogeographic studies related to paleoenvironmental changes.In that context, the case of the Persian ironwood (Parrotia persica (DC.)C.A.Mey) is of particular importance.P. persica (Hamamelidaceae) is one of the most noteworthy endemic relict tree species growing in the Hyrcanian forest (northern Iran and eastern Azerbaijan) (e.g.Yosefzadeh et al., 2010;Sattarian et al., 2011;Sefidi et al., 2011;Ramezani et al., 2013;Adroit et al., 2018Adroit et al., , 2020)).This forest refugium was recently declared a UNESCO World Heritage site due to its unique biodiversity including relic Arcto-Tertiary species such as P. persica (UNESCO, 2019).From a larger temporal perspective, the Persian ironwood has not been restricted to this geographic area and is well known from various Eurasian Cenozoic leaf and pollen records (e.g.Naud and Suc, 1975;Leroy and Roiron, 1996).The oldest European Parrotia pollen records date to the earliest Eocene (cf.Popescu et al., 2021, supplementary data) and fossil leaf records to the Early Miocene (Kvaček, 1998).The intensification of colder conditions in the late Cenozoic had a significant impact on the distribution and isolation of numerous Arcto-Tertiary floral elements (e.g.Suc and Popescu, 2005;DeChaine and Martin, 2006;Milne, 2006;Sagheb-Talebi et al., 2014), including Parrotia persica (Bińka et al., 2003;Jiménez-Moreno et al., 2010;Biltekin et al., 2015).Parrotia is a good example of genus distribution and speciation affected by glaciations, evidenced by the recent discovery of Parrotia subaequalis in mountain forests of Southeast China (Fig. 1) (Fang et al., 1997;Li et al., 1997Li et al., , 2012;;Zhang et al., 2018a).The late Cenozoic isolation of Parrotia populations led to differences in leaf morphology between the two extant species, Iran's P. persica and China's P. subaequalis (Plates I-II).It is thus important to also compare the pollen morphology of both species to note the impact of biogeographical isolation on their complete morphological traits.The similarities between the pollen of both taxa and the diagnostic features that segregate them are important to unravel the origin and evolution of these two species using the palynological record.
This study aimed to investigate whether morphological differences could be detected that segregate P. persica pollen from P. subaequalis.The combined light and scanning electron microscopy based pollen morphology of each taxon was described, measured in detail and compared using acknowledged statistical approaches and morphologically based principal component analyses to establish unbiased results.Based on the morphological traits of extant Parrotia pollen observed with LM, previously illustrated Eurasian fossil pollen grains were remeasured and compared to the range noted for pollen from the two extant species.The detected change in pollen morphology over time is further discussed in relation to both climatic as well as biogeographic evolution in both time and space throughout the Cenozoic.

Origin and accessibility of samples
Pollen from both extant Parrotia, freshly collected as well as herbarium material was analyzed.Flowers were collected from trees occurring in their natural habitats, Parrotia persica flowers from plants in the Hyrcanian forest of Iran, and P. subaequalis flowers from plants in the Yixing forest of eastern China.The herbarium material used for this study belongs to the following collections: 1) Herbarium of the Department of Paleontology of the University of Vienna (HDP-WU); 2) IMBE palynological database, CNRS, Aix-Marseille University (MEPRC); 3) ISEM reference palynological database, CNRS, University of Montpellier.

Sample preparations
The anthers were extracted from the flowers and acetolyzed following the protocol called 'fast way' of Halbritter et al. (2018, p. 103).Individual grains were then photographed with light microscope (LM) and transported onto scanning electronic microscope (SEM) stubs to be photographed following the "single-grain method" as modified by Halbritter et al. (2018, p. 121).

Pollen observations, descriptions, and measurements
Pollen grains were observed with both LM and SEM and described and measured following Punt et al. (2007) and Halbritter et al. (2018).The morphometric characteristics of P. persica and P. subaequalis were established based on measurements with LM from 120 pollen grains: 60 for each species, including 30 pollen grains measured in equatorial view and 30 in polar view (detailed measures are available in Suppl.1).

Statistical analyses of pollen features
Due to previously observed differences in lumen size in P. persica pollen leading to the assumption of some polymorphism (Naud and Suc, 1975;Bińka et al., 2003;Paridari et al., 2012), and due to the larger lumen size in P. subaequalis pollen, a morphometric study was performed on pollen grains in both equatorial and polar views.Using a calibrated square (side = 10 μm) sketched on pollen photographs at magnification ×1000, the lumina were counted and thirty lumina measured in equatorial and polar views for each species.The morphometric characterization of P. persica and P. subaequalis pollen was assessed using complementary multivariate statistical tools based on the six quantitative descriptors (detailed measures are available in Supplement 1).All analyzes were performed using PAST v. 4.02 (Hammer et al., 2001).As a preliminary step, these 6 quantitative descriptors measured for each of the 120 pollen grains included in the analyzed dataset were first transformed in order to optimize within-species multivariate normality (evaluated using the Doornik-Hansen test; (Doornik and Hansen, 1994) and homoscedasticity (i.e., homogeneity of variances and co-variances, as evaluated using the Box's M test; (Rencher and Christensen, 2012) (Tables 1 and 2).The linear measurements (E, Lll, Wll, Lsl and Wsl, see caption Table 1) were log10-transformed, and the Nlumen (see caption Table 1) was square root-transformed, as customarily done for linear and count measurements, respectively (Sokal and Rohlf, 1995).(Sefidi et al., 2011;Zhang et al., 2018b;Liu et al., 2021).Geographic background adjusted from: The World Factbook, Central Intelligence Agency.(For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.)

Principal component analysis
The multivariate distribution of pollen grains measured in equatorial and polar views was first explored separately for each species through correlation-based principal component analysis (PCA; Legendre and Legendre, 2012).Differences between the two views were further investigated through one-way Multiple Analysis of Variances (MANOVA; Rencher and Christensen, 2012) and its nonparametric equivalent, one-way PerMANOVA based on the Mahalanobis generalized distance matrix, 99.999 permutations; (Anderson, 2001).These tests were coupled with two-group linear discriminant analysis (LDA; Legendre and Legendre, 2012) in order to visualize group differences along the canonical axis tested by MANOVA.Additionally, the multivariate difference of the two species in both equatorial and polar views was evaluated using two-group one-way MANOVA (two species) and fourgroup one-way MANOVA (two species × two views).The same twogroup and four-groups were conducted on a one-way and two-way Mahalanobis-based PerMANOVA coupled with the corresponding LDA.

Climatic and biome niche analysis
We compiled historical climate data  from 196 georeferenced occurrences of P. persica and seven of P. subaequalis using WorldClim vers.2.1 (https://www.worldclim.org/data/worldclim21.html)at a resolution of 30 s (ca.1km 2 ; Fick and Hijmans, 2017; see Supplementary File 2).The georeferenced data were filtered so that multiple occurrences with the same coordinates were treated as a single data point.To characterize climatic and ecological niches of the two modern species, we created distribution maps for each species using the GBIF dataset (https://www.gbif.org/).These occurrences were plotted onto 5 arc minutes grid Köppen-Geiger climate maps (1986-2010 data;Kottek et al., 2006;Rubel et al., 2017) to establish "Köppen signatures" for both species; and on major terrestrial biome maps (Olson et al., 2001) to assess their biome preferences.For the Köppen-Geiger plots, the georeferenced datasets were filtered so that multiple occurrences in a single grid cell were only counted once (labeled 'unique grid cells' in the diagrams; Supplementary File 3).Likewise, for the biome plots, georeferenced data were filtered so that multiple occurrences within the same coordinates were treated as single data points (labeled 'unique localities' in the diagrams; Supplementary File 3).Georeferenced data and the Köppen-Geiger maps with a 5 arc minutes resolution were processed using the "Sample Raster Values" Toolbox in QGIS Version 3.16.4-Hannover.The biome shape files were processed using the "Geoprocessing Tool" intersection in QGIS.The biomes and Köppen-Geiger climates occupied by both Parrotia species are shown as maps generated with QGIS and as frequency (proportional distribution) diagrams (Supplementary File 3). Ecel files provide the raw point data (Supplementary File 2).

Statistical approach and pollen comparison
The correlation-based PCA achieved for each species show an overall good superposition of pollen grains measured in equatorial and polar views in the first principal plane (explaining 61.7% and 67% of the total variability for P. persica and P. subaequalis, respectively; Fig. 2A, C).In both cases, the multivariate space appears to be structured in the same way, with the lumen density vs. size of the smallest and largest lumina driving the first principal component, and the equatorial diameter driving the second principal component.Parametric and nonparametric MANOVAs confirm for each species the impossibility of distinguishing between specimens observed in the equatorial view and those observed in the polar view (Table 1).The significant result obtained for P. persica is a spurious by-product of the presence in the analyzed sample of 22 (37%) large-E pollen grains (equatorial diameter) pollen grains coming from a single microscope slide (n°35,249) and involving 13/22 (59%) grains in polar view, while the small-E group involves 17/38 (44.7%) grains in polar view.Similar tests achieved without the equatorial diameter descriptor return non-significant results confirming the morphometric homogeneity between equatorial and polar views (MANOVA: λ = 0.86, p = 0.13 NS; PerMANOVA: F = 1.69, p = 0.13 NS).Finally, the two-group LDA (Linear Discriminant Analysis) illustrates the lack of morphometric differences between equatorial and polar views for each species (Fig. 2B, D).
On the contrary, compared to each other, pollen from the two Parrotia species show highly significant morphometric differences based on pooled and separate equatorial and polar views (Table 2).The two-way PerMANOVA (species × view) confirms that most of the morphometric heterogeneities between groups come from differences between species, with no significant interaction between species and view factors.The corresponding LDAs also illustrate significant differences, showing that the morphometric difference between the two species is mainly driven by lumen density vs. size of the smallest and largest lumina (Fig. 3).

Extant Parrotia biomes and climate niches
Both extant Parrotia species are part of the "Temperate Broadleaf and Mixed Forests" biome (Olson et al., 2001;Supplementary File 3).Based on our georeferenced dataset, both species display relatively similar mean annual temperature (MAT) preferences and coldest month mean temperature (CMMT) tolerances (Supplementary Files 2 and 3).Significant differences were observed with precipitation.While P. subaequalis thrives under fully humid conditions (Cfa climate; Fig. 4), with 1052 (1287) 1585 mm mean annual precipitation (MAP) and 485 (549) 666 mm of precipitation during the three warmest month, P. persica tolerates 291 (750) 1195 mm MAP, with strong seasonality and a precipitation depression during the summer months.With only 22 (76) 135 mm of precipitation during the three warmest months, P. persica thrives mainly under a Csa climate (Fig. 4), bordering in the east of its natural distribution to BS climate and extending at higher elevations along the mountain ranges near the Caspian Sea into Cfa climate.

Concrete distinction of pollen
The pollen of both Parrotia species is easily identified at the genus level.The suspected polymorphism of P. persica pollen can be rejected based on the LDA, revealing that the P. persica pollen is morphometrically as homogeneous as those of P. subaequalis.When compared, the pollen grains of P. persica are slightly larger and their exine is thicker than those of P. subaequalis (Plates III and IV; Tables 1 and 2).Although pollen grains of both species are reticulate in LM, the lumen density is much finer in P. persica than in P. subaequalis.These differences are more visible with SEM observations, where the sculpture of P. persica is micro-reticulate and the sculpture of P. subaequalis is reticulate (compare Plate III, 3 and 4 to Plate IV, 3 and 4).SEM further differentiates the reticulate sculpture of the two species; the muri are broader, the diameter of the lumina is larger, and the number of free-standing columellae Plate V. Parrotia persica C.A.Mey, modern pollen grain (slide 20407 of the ISEM collection, plant specimen collected in Iran).1-6, Pollen in equatorial view (aperture facing), L.O. analysis: 1, surface of reticulate ornamentation; 2, base of lumina; 3, foot of columellae; 4-5, colpi; 6, optical section.7-12, Pollen in polar view, L.O. analysis: 7, surface of reticulate ornamentation; 8, base of lumina; 9, foot of columellae; 10-11, colpi; 12, optical section.13-18, Pollen in equatorial view (aperture facing), L.O. analysis: 13, surface of reticulate ornamentation; 14, base of lumina; 15, foot of columellae; 16-17, colpi; 18, optical section.19-24, Pollen in equatorial view (intercolpium facing), L.O. analysis: 19, surface of reticulate ornamentation; 20, base of lumina; 21, foot of columellae; 22-23, colpi; 24, optical section.25-30, Pollen in polar view, L.O. analysis: 25, surface of reticulate ornamentation; 26, lumina; 27, base of lumina; 28, foot of columellae; 29, colpi; 30, optical section.Scale bar = 10 μm. is higher in P. subaequalis compared to P. persica.Furthermore, sculpture elements along the colpus margins are differently arranged in the two species (compare Plate III, 5 and 6 to Plate IV, 5 and 6).Furthermore, statistical approaches confirm the significant difference between pollen of the two Parrotia species.Combining the available evidence, P. persica is best characterized by a dense and small lumina, while P. subaequalis is characterized by a less dense and larger lumina, as summarized in Table 1.Ultimately, the cross-validation of the resulting two-species linear discriminant classifier shows that based on this reference dataset, (sub)fossil pollen grains can now be confidently assigned to either one of the Parrotia species with an 85.8% correct-assignment rate.

Inputs for paleopalynology and affiliation to extant species
The equatorial diameter (E) and length of the polar axis (P) in fossil Parrotia pollen is usually not comparable to that of pollen from extant Parrotia due to the hydrated state of extant material studied with LM.Potential fossil Parrotia pollen is mostly preserved in dehydrated state, either with the aperture enfolded or widely open (lacking colpus membrane).Fossil pollen grains with enfolded apertures are stretched along their polar axis, therefore, they have an excessively long polar axis compared to their equatorial diameter.Grains with widely open apertures are collapsed, and their poles are compressed; therefore, they have an excessively wide equatorial diameter.This makes it hard to compare the size/outlines/shape (e.g. the P/E-ratio) of single dispersed fossil Parrotia pollen to that of either P. persica or P. subaequalis.Irrelevant to this is the sculpture observed in LM and/or SEM.The number and size of the lumina are not affected by the P/E ratio and can be measured in both fossil and extant pollen and compared.As evident in Table 1, there is a clear difference in the lumen size range and partly in their number per 10μm 2 between the two living species.These features currently seem to be the only tool for possibly segregating fossil Parrotia pollen and investigating if they are morphologically/taxonomically closer to one or the other extant species.To test this, we measured the lumen features from potential fossil dispersed Parrotia pollen previously described from Late Oligocene to Pliocene localities of Europe (see Table 3, and references therein).Interestingly, the oldest European records from our comparison, which are of late Oligocene to Middle Miocene age, suggest affiliation to P. persica.The younger records, of Late Miocene to Pliocene age, suggest affiliation to P. subaequalis.The reasons for this might be as follows: 1) the pollen type of P. persica is the basal/ ancestral pollen within Parrotia, and European Oligocene to Middle Miocene Parrotia plants, one species or more, produced pollen comparable to that of modern-day P. persica.The change or addition in morphology could reflect a divergence within the genus and the origin of a new species that is now confined to East Asia; 2) this has no meaning and the affiliation 'older versus younger records' to either of the living species might change and intertwine back in time when a larger number of records are considered.I any case, such assumptions would have to be supported or rebutted in a future more detailed study based on a larger fossil pollen collection from different Eurasian locations.

Paleoecological and paleoclimatological considerations
Since plant species of the same genus can characterize drastically different ecological niches, it is important to be able to segregate tree pollen at intrageneric levels, especially when those taxa are also present in the fossil record.For example, the genus Acer comprises more than hundred different plant species developing within drastic different environments.In western Eurasia (Europe, Caucasus), pollen cannot be used to discriminate between the drought-adapted Acer monspessulanum L., Acer campestre L., Acer cappadocicum Gled., which occupy a closed montane forest even up to a mesic valley (Akhani, 1998), and the humid forest species such as Acer velutinum Boiss., Acer pseudoplatanus L., Acer platanoides L. (Beug, 2004, p. 250).Another example is Quercus, also very widespread in the world and in the fossil record.The Quercus robur-pubescens pollen type cannot be discriminated between various white oaks that can be swamp forest or dry steppe forest elements (Beug, 2004, p. 144).However, the possibility to discriminate the evergreen (e.g., Quercus ilex-type) from deciduous oak (Quercus robur-type) has been an asset for paleoecologists to understand the ecological history of Mediterranean forest ecosystems.
As evident herein, Parrotia is one of those genera in which it is possible to differentiate between the pollen of its species.The case of Parrotia is interesting because this Arcto-Tertiary floral element can be useful in determining Quaternary interglacial paleo-refugia.So far, Parrotia pollen grains described from the Cenozoic of Eurasia have mostly been attributed to P. persica (Stachurska et al., 1973;Naud and Suc, 1975;Leroy and Roiron, 1996;Jiménez-Moreno et al., 2007;Jiménez-Moreno and Suc, 2007;Suan et al., 2017;Suc et al., 2020;Popescu et al., 2021).The present study shows that previous fossil pollen records could potentially be revised based on the pollen morphology and that future studies comprising fossil Parrotia pollen need to consider the morphological differences between the two extant species of the genus.
The main reasons why it is important to revise previous fossil Parrotia pollen records and correctly affiliate new finds is the potential impact on paleoenvironmental interpretations.Today, P. persica and P. subaequalis do not share the same bioclimatic conditions.Although their biomes can both be defined as Temperate Broadleaf & Mixed Forests, the climate sustaining P. subaequalis is a fully humid warm temperate climate with hot summers, a typical Cfa climate (e.g., Kottek et al., 2006;Fig. 4), while in the Hyrcanian forest, P. persica thrives under a warm temperate climate with hot but dry summers, a typical Csa climate (e.g., Kottek et al., 2006;Fig. 4, Supplementary Files 2 and 3).This difference in the present climatic conditions is also supported by the Computerized Bioclimatic Maps of the World (Rivas Martinez et al., 2011) where P. persica is shared between the bioclimate n°33  (i.e.Temperate-oceanic) and n°26 (i.e.Mediterranean-xeric-oceanic) while P. subaequalis remains in n°32 (i.e.Temperate-continental) bioclimatic conditions.Unfortunately, the allegedly narrow ecological niche of P. persica, is commonly used to infer paleoclimate conditions (temperatures and precipitations) for Eurasian paleoenvironments (Emberger and Sabeti, 1962;Ramezani et al., 2013).But in their publication on the palynoflora from the Lavanttal Basin, Grímsson et al. (2015) referred to several studies focused on P. persica providing significant range in the amount of for example annual precipitation, mean annual temperature, elevation, and associated plant species.
Overall, a comparative ecological niche modeling of both extant species is needed in order to put forward a more reliable paleoenvironmental interpretations based on affiliation to either of the extant species.Moreover, because of the restricted distribution of extant Parrotia some climatic parameters could be underestimated (e.g.Bruch et al., 2002;Uhl et al., 2007;Thiel et al., 2012).

Conclusions
Comparison between pollen grains of Parrotia persica (from the Hyrcanian forest of Iran and Azerbaijan) and P. subaequalis (from temperate forests of south-eastern China) demonstrates significant morphological differences.The two-species LDA ultimately provides a linear discriminant classifier so dispersed pollen grains can be assigned to either species based on six analyzed descriptors.This distinction can be done using simple light microscopy; P. subaequalis is characterized by large and sparse lumina, and P. persica is best identified by small and dense lumina.Until now, palaeopalynological studies, affiliated fossil Parrotia type pollen automatically as being grains of P. persica.The recent discovery of the sibling P. subaequalis and now, with our study, the demonstration of its significant difference in pollen morphology when compared to P. persica led to re-evaluate the former palaeoenvironmental estimations, especially in Europe where the fossil records between Late Miocene to Pliocene support an affiliation to P. subaequalis pollen type.Such observation could suggest a divergence within Parrotia and the emergence of the new species, that is now restricted to Eastern Asia.

Declaration of Competing Interest
The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.

Fig. 1 .
Fig. 1.Geographic maps showing the distribution of Parrotia persica in Iran and P. subaequalis in China.Distribution of extant populations marked with red based on(Sefidi et al., 2011;Zhang et al., 2018b;Liu et al., 2021).Geographic background adjusted from: The World Factbook, Central Intelligence Agency.(For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.) Plate I. Parrotia persica in its natural habitat in the Hyrcanian forest, Iran.(1) Fruits, usually less than 1 cm in diameter.(2) Flowers, usually 1 cm long.(3-4) Leaves, oblong to obovate, commonly 10 cm in length, can reach 15 cm.(5-6) Tree or sometimes large shrub, usually 8-10 m tall, but can be up to 25 m.Scale bar = 1 cm in 1-4; 1 m in 5 and 6.Photos (1) from Dr. Hamid Gholizadeh; (2-4) from the open access website ebben.nl.

Fig. 2 .
Fig. 2. Principal Component and Linear Discriminant Analyses of Parrotia pollen.Correlation-based Principal Component Analyses (A, C) and two-group (equatorial view vs. polar view) Linear Discriminant Analyses (B, D) of 60 pollen grains from Parrotia persica (A, B) and P. subaequalis (C, D).
μm 2 of pollen surface, which corresponds to the lumen density in fine.Lll and Wll = length and width of the largest lumen observed in the same 10 μm 2 area.Lsl and Wsl = length and width of the smallest lumen observed in the same 10 μm 2 area.Measurements from original publications or/and based on illustrated fossil pollen.

Table 1
Light microscopy measurements of Parrotia pollen.Note: Statistical summary of the six quantitative descriptors measured for P. persica and P. subaequalis in equatorial and polar views (see Suppl. 1 for full detailed values).E (μm) = equatorial diameter of pollen.Nlumen = number of lumina per 10 μm 2 of pollen surface, which corresponds to the lumen density in fine.Lll and Wll = length and width of the largest lumen observed in the same 10 μm 2 area.Lsl and Wsl = length and width of the smallest lumen observed in the same 10 μm 2 area.

Table 2
Multivariate statistical tests and morphometrical characterization of Parrotia pollen.