Phytosociological and ecological discrimination of Mediterranean cypress (Cupressus sempervirens) communities in Crete (Greece) by means of pollen analysis

Sixty modern surface samples collected from mosses in different cypress forest communities (Cupressus sempervirens L.) on the island of Crete (Greece) were analysed for their pollen content. The samples were taken from six different cypress phytosociological associations between 23 and 1600 m asl, and fall within distinct rainfall and temperature regimes. The aims of this paper are to provide new data on the modern pollen rain from the Aegean islands, and to perform these data using multivariate statistics (hierarchical cluster analysis and canonical correspondence analysis) and pollen percentages. The discrimination of pollen assemblages corresponds to a large extent to the floristic differentiation of Cupressus sempervirens forest vegetation and indicates the existence of three new associations.


Introduction
The Mediterranean, common, or Italian cypress (Cupressus sempervirens L.) is a medium-size evergreen coniferous tree whose natural habitats are found in the semi-arid mountains around the eastern Mediterranean basin and the Middle East. It has been widely cultivated and naturalized elsewhere since historical periods and its natural range is unclear (Eckenwalder, 2009;Caudullo & De Rigo, 2016). Natural stands occur in the eastern Mediterranean basin over several geographically non-adjacent areas reaching eastwards the Caucasus and south-western Iran, from sea level (Crete) up to 2000 m asl in Turkey (Intini & Della Rocca, 2004). Its natural distribution comprises the Aegean islands (Greece), Cyprus, Turkey (south Anatolia), northeast Africa (Libya, Tunisia), and the Middle East (Iran, Jordan, Lebanon and Syria) (Farjon & Filer, 2013), where cypress occurs mostly as disjunct populations (Zohary, 1973;Roques et al., 1999). Two morphological varieties of cypress are known (Tutin et al., 1993): the wild type var. horizontalis (Miller) Aiton, which is widely represented by native populations and is characterized by a broad conical crown with the main branches forming a wide angle with the trunk; while the fastigiated var. pyramidalis Nymann (= var. stricta Aiton) is the most widely planted and cultivated since ancient times for its columnar and dense crown with main branches growing upwards close to the trunk (Zacharis, 1977;Farjon, 2010).
Genetic and palaeobotanical data (Papageorgiou et al., 1993(Papageorgiou et al., , 1994(Papageorgiou et al., , 2005Korol et al., 1997;Bagnoli et al., 2009;Manescu et al., 2011;Velitzelos et al., 2014) point out: i) high genetic differentiation between natural and domesticated cypress populations; ii) the reduction in allelic richness in Italian and Tunisian populations due to genetic drift originating from a long tradition of cypress cultivation and/ or a drastic reduction in effective population sizes of natural cypress since the Tertiary; and, iii) that the present natural populations of Mediterranean cypress are just the remnants of an extensive cypress forest of the Pliocene, distributed throughout the central and eastern Mediterranean areas, which was progressively disappearing as a consequence of human activities (irrational felling, grazing and wildfires) and canker disease caused by Seridium cardinale (Wag.) Sutton & Gibson, particularly in the island of Crete (Papanastasis et al., 1990;Xenopoulos et al., 1990;Rackham & Moody, 1996). In Greece, the Mediterranean cypress is native only in Crete and some of the East Aegean islands (Samos, Rhodes, Kos, Symi, Milos, Chios, Kalymnos; perhaps also in Thasos and Samothraki) (Papageorgiou et al., 2005;Brofas et al., 2006;Dimopoulos et al., 2013). The Greek Cupressus sempervirens woods are included in the Natura 2000 network of Protected Natural Areas of Greece (Habitat 9290) and they are considered of high conservation value because they comprise an endemic and rare habitat type, which is important for endemic species and it faces several threats (Papastergiadou et al., 1997;Dafis et al., 2001;Dimopoulos et al., 2006). Cupressus sempervirens is a pioneer species, growing quickly on most types of soils, including rocky and compact ones, such as calcareous, clayish, dry and poor soils (Xenopoulos et al., 1990). It is adapted to the Mediterranean climate with dry and hot summers and rainy winters, as well as to the semi-arid climate in the eastern and interior areas of its range (Farjon, 2010). The Mediterranean cypress is able to tolerate prolonged drought and high temperatures. Particularly, the var. horizontalis is resistant to ignition because of the high ash content and the ability of the leaves to maintain a high water content during the summer (Della Rocca et al., 2015).
Although Cretan cypress forests have been relatively well studied from a phytosociological point of view, there are still a number of controversies among researchers as well as confusing syntaxonomic assignments. A first approach was done by Zohary & Orshan (1965), who described the Cupresseto-Aceretum orientalis association and its corresponding cupressetosum subassociation, also previously informally described by Rechinger & Rechinger-Moser (1951), limited to the highest altitudes. Later, Barbero & Quézel (1980) corrected the illegitimate nomenclature of Zohary & Orshan (1965), recognizing the new association Luzulo nodulosae-Cupressetum sempervirentis as the most widespread in the island between 1000−1100 m asl and even constituting the tree line over 1600−1700 m asl. They recognized three subassociations: the aforementioned cupressetosum in the supra-mediterranean bioclimatic belt, and the two new pinetosum brutiae (meso-mediterranean) and berberidetosum creticae (montane mediterranean belt).
In a later work, concerning orophilous communities from the central and eastern Mediterranean area, Brullo et al. (2001) described two new associations of Cupressus sempervirens woodlands with very open canopy in the Lefka Ori range: the first one in the lower and middle montane mediterranean belts at 1000−1400 m asl (Daphno sericeae-Cupressetum sempervirentis), and the second one at the timberline (1400-1800 m asl) in the upper montane to oro-mediterranean belts (Junipero oxycedri-Cupressetum sempervirentis). These two new communities have been considered by Bauer & Bergmeier (2011) as synonymous with the Luzulo nodulosae-Cupressetum sempervirentis association and the Luzulo nodulosae-Cupressetum sempervirentis berberidetosum creticae subassociation, respectively. Brullo et al. (2001) situate both new associations within the Pino-Juniperetea class due to the abundance of orophilous shrubs. More recently, Brofas et al. (2006) described five communities characterized by the constant presence of Cupressus sempervirens in combination with other tree or shrub species: Quercus coccifera, Q. ilex, Erica arborea, Acer sempervirens and Pistacia lentiscus. The former, with kermes oak, was found between the thermo-and the lower supra-mediterranean bioclimatic belts (120−800 m asl) and occasionally in the upper supra-mediterranean and montane mediterranean ones (850−1240 m asl) on limestone, marly conglomerates, schists or, more rarely flyschs. The one with lentisk occurs mainly in the thermo-mediterranean belt (10−300/350 m asl) and rarely extend to the mesomediterranean one (250−550 m asl) on similar soil types. They also documented in their relevés the Luzulo nodulosae-Cupressetum sempervirentis association, but they were not able to distinguish the three subassociations described by Barbero & Quézel (1980); although they differentiate this association from the community formed by the cypress and Acer sempervirens precisely by the presence of the latter species in the lower and middle montane mediterranean belts (1100−1500 m asl) on hard limestone rocks and occasionally in the supra-mediterranean belt at 800 m asl on schists and even in the upper montane and oro-mediterranean ones (1500−1800 m asl) also on limestone. The authors consider the Acer sempervirens-Cupressus sempervirens community as a high altitude endemic woodland association restricted to Crete, although they do not clasiffy it in a certain syntaxonomical rank. In fact, from a bioclimatic and floristic point of view this community resembles the Daphno sericeae-Cupressetum sempervirentis association described by Brullo et al. (2001). Two other cypress communities cited by Brofas et al. (2006) with Quercus ilex and Erica arborea also occupies the thermo-and the meso-mediterranean bioclimatic belts (200−550 m asl), growing on relatively moist sites.
Finally, in the most recent work we know to deal with cypress communities, Bauer & Bergmeier (2011) make a detailed review of cypress mountain woodlands of western Crete above 500 m asl, although the authors do not use Brullo et al. (2001) andBrofas et al. (2006) relevés because they are considered incomplete. They cite the rare presence of cypress in relevés of the Aceri sempervirentis-Quercetum calliprini association in the Psiloritis range (900−1385 m asl), i.e. in mixed woodlands invariably composed by Quercus coccifera and Phillyrea latifolia, where the presence of Quercus ilex and Acer sempervirens is also occasional. Some inventories of this association resemble floristically those of the Quercus coccifera-Cupressus sempervirens community described by Brofas et al. (2006) that occasionally reachs the montane mediterranean belt. Bauer & Bergmeier (2011) also recognize in their relevés the Luzulo nodulosae-Cupressetum sempervirentis cupressetosum subassociation between the meso-and the supra-mediterranean belts (400-1070 m asl), which, unlike indicated by Brofas et al. (2006), incorporates Acer sempervirens but also Pinus brutia. They also recognize the two other subassociations described by Barbero & Quézel (1980): berberidetosum creticae (830-1480 m asl) and pinetosum brutiae (150-1075 m asl). Bauer & Bergmeier (2011) point out that there is a gradual transition between Pinus brutia and Cupressus sempervirens forests without obvious ecological evidence (Rechinger & Rechinger-Moser, 1951). In fact, pine forests other than included in the pinetosum brutiae subassociation occur on the lower slopes of the southern foothills of Crete. Among these, Barbero & Quézel (1980) described the Irido cretensis-Pinetum brutiae community for thermophilous woodlands located at lower altitudes (300−800 m asl), which often incorporate isolated Cupressus sempervirens trees.
In this paper, we study modern pollen samples from Crete to discriminate Cupressus sempervirens forest communities, that is, the exploration of how well local vegetation is represented in pollen samples-assemblages. The advantage of modern pollen rain studies is that cushion mosses collect the airborne pollen of the last 5−30 years (Pardoe et al., 2010), while the floristic inventories depend intrinsically on the season of the year in which they are made. In this sense, multivariate analyses are a powerful tool in the comparison and diagnosis of different plant communities from their pollen content in Mediterranean European countries (Court-Picon et al., 2006;Mazier et al., 2006;López-Sáez et al., 2010Glais et al., 2016). The main objectives of our study are (i) to explore how contemporary vegetation is depicted in surface pollen assemblages, (ii) to identify pollen indicators of Cupressus sempervirens forest syntaxa, and, (ii) to establish statistic relationships between pollen rain and vegetation patterns in order to use them for paleoecological reconstruction in future studies.

Study area
The island of Crete is located south of the Aegean Sea in the southeastern part of the Mediterranean region ( Figure 1). With 8336 km 2 (the maximum length is 269 km and the maximum width 60 km) is the largest island of Greece and the fifth largest island of the Mediterranean (Legakis & Kypriotakis, 1994). It is an extremely mountainous territory with four main massifs from west to east: Lefka Ori or White Mountains range (Pahnes: 2453 m asl), Psiloritis or Ida range (Timios Stravos: 2456 m asl), Asterousia range (Kofinas: 1231 m asl) and Dikti range (Spathí: 2148 m asl). Various calcarerous rocks (limestones and dolomites) dominate the mountain areas, whereas Neogene sediments (limestones, sandstones and marls) cover large areas of the lowlands. There are also ortho-quartzites, phyllites, flyschs, Quaternary rocks and alluvian deposits (Higgins & Higgins, 1996). from west (2118 mm on the Askifou upland, 740 m asl) to east (440 mm on the plain of Ierapetra, 10 m asl) and from north (Mediterranean pluviseasonal oceanic) The Cretan climate is typically Mediterranean with hot and dry summers, and mild and wet winters. The mean annual rainfall is about 750 mm, and decreases to south (Mediterranean xeric oceanic), but increases with altitude (Naoum & Tsanis, 2004;Koutroulis et al., 2011). There is also a slight increase (1-1.5ºC) in mean annual temperature from northwest to southeast and a decrease with altitude. The eastern and southern parts of the island are more arid than the western and northern ones, as there is higher precipitation in the northwestern coastal areas and lower in the southeastern part of the island (Chartzoulakis & Psarras, 2005). Crete belongs to the Cretan biogeogaphical subprovince, Graeco-Aegean province, Eastern Mediterranean subregion, Mediterranean region (Rivas-Martínez et al., 2004).

Field methods -vegetation and pollen sampling
Vegetation and pollen samples were collected in 2017 at 60 locations on Crete ( Figure 1). Sampling sites were chosen to cover a certain variety of vegetation types in the whole island of Crete, in a wide altitudinal range, although the aim was mainly to analyse modern pollen rain in order to sample plant communities dominated or co-dominated by Cupressus sempervirens. Moss polsters samples were collected at each location to provide modern pollen data, with positional and altitudinal data recorded using a portable Juno 3D Trimble Ltd. Global Positioning System (GPS) device. Moss samples were collected over an area of approximately 100 m 2 by taking multiple moss polsters from the concerned site to ensure an even representation, following Sugita (1994) and Hicks et al. (2001) recommendations. The subsamples were sealed in plastic bags and mixed into one sample per site. A relevé of vegetation was also made at each sampling site following the physotociological approach (Braun-Blanquet, 1979). Ten environmental variables were available for 60 sites ( Table 1). Six of these were bioclimatic variables obtained from the WorldClim database (Fick & Hijmans, 2017) in a 30-sec resolution (approximately 1 km 2 ), while the other four were tree cover, altitude, bedrock type and grazing pressure. Tree cover was graded on an ordinal scale from 1 to 5 at each sampling point (100 m 2 ) as follows: 5 (75-100%), 4 (50-75%), 3 (25-50%), 2 (5-25%), 1 (>0-5%). Bedrock types were obtained from the geological map of Greece at a scale of 1: 500,000 (I.G.M.E., 1983) and confirmed during the field study. Grazing pressure was estimated on a scale of 0 to 4 (Court- Picon et al., 2006).

Laboratory methods -pollen
Moss polsters samples of approximately 10 cm 3 were homogenized prior to extraction. The samples were sieved through 1 mm screens to remove larger particles (e.g., leaves, twigs, and gravel) and then processed following the standard protocol developed by Faegri & Iversen (1989). Samples were stored in glycerol, mounted on microscope slides and examined with a Nikon Eclipse 50i light-microscope (Melville, NY, U.S.A.) to identify pollen, spores and nonpollen palynomorphs. Routine counting was carried out at 400x magnification. Pollen grains, spores and non-pollen palynomorphs were identified according to Moore et al. (1991) and López-Sáez & López-Merino (2007) at the lowest currently possible taxonomical level. "Type" groups of several taxa that are morphologically indistinguishable were used. This has been the case of Cupressaceae species present in Crete (Cupressus sempervirens, Juniperus sp.), as they share the same characteristics (stenopalynous) under the light microscopy (Moore et al., 1991;Kurmann, 1994;Hidalgo et al., 1999). Erica arborea and Erica manipuliflora pollen types were palynologically discriminated according to the Pal-Dat Palynological Database (www.paldat.org). Erica arborea-type was defined as small-size tetrahedral tetrads (pollen unit 10-25 µm) of spheroidal shape; while Erica manipuliflora-type was defined as medium-size tetrahedral tetrads (pollen unit 26-50 µm) of isodiametric shape. Pistacia lentiscus and P. terebinthus were palynologically identified according to Burgaz et al. (1994). A minimum of 300 pollen grains were identified and counted for each sample. Pollen percentages were calculated using a pollen sum excluding spores and non-pollen palynomorphs, and presented as bars in a pollen diagram. Tilia and TGView (Grimm, 1992) and CorelDraw software were used to plot the pollen diagrams (Figures 2 and 3). The terms 'local' and 'regional' used in the text refer to different pollen source areas according to Prentice (1985).

Statistical analyses
To identify clusters of samples based on their pollen content and hece to define specific Cupressus sempervirens forest communities, we used multivariate analysis. Only palynomorph taxa present at > 3% in at least one of the samples were included. The analyses were performed on recalculated percentages after all modifications had been made. Classification was performed by hierarchical cluster analysis (HCA) using the matrix of the Euclidean distance and Ward's minimum variance method (Ward, 1963) with software PAST (Hammer et al., 2001). The percentage values of the taxa were not transformed. The hierarchical relationships between clusters are illustrated by the dendrogram in Figure 4. Later, we used ordination analyses to estimate the relationships between environmental variables, pollen assemblages and palynomorph types, eliminating those samples that the HCA clearly differentiated as not belonging to cypress forests (clusters Aa1 and Ba2b). Canonical correspondence analysis (CCA) was used as a unimodal interpretation method, because a previously applied detrended correspondence analysis (DCA) pointed to a unimodal response (gradient length > 2 standard deviation of species turnover units) of palynomorph types (variables) instead of lineal responses of taxa (ter Braak & Prentice, 1988). CCA was performed using square-root transformation of the percentage of palynomorph taxa and down-weighting of rare taxa ( Figures 5 and 6) with CANOCO version 4.5 (ter Braak & Šmilauer, 2002). Each environmental variable was entered separately into the analysis and its significance was assessed using the Monte Carlo permutation test with 999 permutations. The analysis was run with scaling for inter-sample distances to relate the gradient in pollen assemblages to explanatory variables. Forward selection of explanatory variables was used to provide a ranking of the importance of specific variables and to avoid co-linearity (ter Braak & Šmilauer, 2012). The results are presented in Tables 2 and 3. Correlations between CCA axes and environmental variables were calculated using the non-parametric Kendall coefficient in Statistica 9.1 software (http://www.statsoft.com) and are presented in Table 4.

Results
The pollen and non-pollen palynomorph percentage data of the 60 modern surface samples ( Figure 1, Table  1) are summarized in two pollen diagrams (Figures 2  and 3). A total of 67 pollen and non-pollen palynomorph taxa were identified during the analysis and counts of the 60 surface samples collected. The 60 modern surface pollen samples (hereinafter samples) were classified by means of HCA into twelve sample groups, which correspond to ecologically and floristically interpretable vegetation units. On the first division level, the dendrogram of the HCA performed on pollen data ( Figure 4) shows a good differentiation between forests dominated by cypress trees (cluster B) and mixed woodland formations of cypress and other species or woodlands hosting scattered cypress trees (cluster A). This division is probably related to high pollen frequencies (> 20%) of Cupressus/Juniperus pollen type in samples from cluster B (Figure 2). The next two clusters within cluster A are distinguished on the basis of high percentages (> 25%) of Quercus ilex/ coccifera (cluster Aa) and Pistacia lentiscus (cluster Ab) respectively ( Figure 3). Inside cluster Aa, we can detach, at a smaller range, the sub-clusters Aa1 and Aa2; respectively separating samples with Cupressus/ Juniperus values lower than 3% (regional origin; subcluster Aa1) and ~15% (local origin; sub-cluster Aa2). Relating to cluster B, the major segregation separates Ba and Bb clades. The main disjunction promoted here is due to the high percentages (> 50%) of Cupressus/ Juniperus pollen type in cluster Bb. The division of cluster Ba produced two main clades (Ba1 and Ba2). The first cluster Ba1 includes samples with Cupressus/ Juniperus values ranging from 25 to 50% (except sample CS10), while in cluster Ba2 samples these values are less than 25%. Inside cluster Ba1, the sub-clusters Ba1a and Ba1b are segregated by the high values (12−38%) of Pinus brutia in the first and the very low (< 1%) in the second one (Figure 2). At a lower range, the subcluster Ba1a is subdivided into three clades as abundant Berberis cretica, Prunus and Rhamnus (Ba1a-1) or phrygana (Sarcopoterium spinosum, Thymbra capitata, Verbascum, Euphorbia acanthothamnos) pollen taxa (Ba1a-2 and 3) (Figure 3). Likewise, the sub-cluster Ba1b is also subdivided into three smaller clades depending on the abundance of Pistacia lentiscus (Ba1b-1), Daphne and Berberis cretica (Ba1b-2) or phyrgana pollen taxa (Ba1b-3). Finally, cluster Bb is subdivided into two main clusters (Bb1 and Bb2), segregated by the abundance of Berberis cretica and Phlomis in the first, as well as Pinus brutia, Euphorbia acanthothamnos, Sarcopoterium spinosum and Thymbra capitata in the second one (Figures 2 and 3). The ordination of 53 sampling sites on the CCA biplot graph is shown in Figure 5, with arrows representing environmental variables, and their length linked to the correlation of the specific variable with the ordination axes. Thus, longer arrows indicate greater importance of the corresponding factor for sampling site variation. The samples corresponding to clusters Aa1 and Ba2b were not included in this analysis since they did not correspond to cypress forests. The first two axes of canonical correspondence analysis (CCA) explained 38.6% of the total pollen taxa variation and the first four axes 47.3% (Table 3). The first axis of the CCA explained almost 54.2% of the pollen taxaenvironment relationship. The eigenvalue (0.256) corresponding to the first canonical axis is highly significant (p < 0.001), indicating significant difference in pollen assemblages among sampling sites. This CCA axis 1 seems to be strongly negatively correlated to altitude, PA, PM and Pm and positively correlated with TA, TM and Tm ( Figure 5, Table 4). Many pollen taxa are discriminated along this axis ( Figure  6); those displaying the highest species scores were typical of highland areas (negative values: e.g. Prunus, Rhamnus, Zelkova, Daphne, Berberis vulgaris, Acer) or maquis vegetation from lowlands (positive values: e.g. Ceratonia siliqua, Prasium, Erica manipuliflora, Pistacia lentiscus, Cistus creticus, Calicotome, Olea europaea, Myrtus). The second axis accounted for about 20% of the overall species-environment relationship (r = 0.741, p < 0.001) and was strongly associated with tree cover with high positive correlation, as well as with grazing pressure with low negative correlation ( Figure  5, Table 4). Pollen taxa strongly associated with this axis are Laurus, Pteridium, Pistacia terebinthus, Styrax, Ruscus, Smilax and Tamus with positive correlation, and Zelkova, Prunus and Rhamnus with negative one (Figure 6). Another variable linked to human activities (e.g. grazing pressure), was also weakly correlated with this axis. The third axis (explaining 9.8% of the pollen taxa-environment relationship) seems to be fairly linked to tree cover and grazing pressure, and to a lesser extent to TA and Tm (Table 4).
Along CCA axis 1 there seems to be an altitudinal distribution of sub-cluster Ba1b samples: sub-cluster Ba1b-1 samples (178−323 m asl) are located at the positive side of the CCA axis 1, negatively correlated with altitude; sub-cluster Ba1b-3 samples (699−739 m asl) are grouped in an intermediate position along CCA axis 1 near zero; while sub-cluster Ba1b-2 samples (1239−1242 m asl) are clearly located on the CCA axis 1 with negative values -as cluster Bb samples-and positively correlated with altitude ( Figure 5). In fact, sub-cluster Ba1b-2 samples are located close to the pollen indicator taxa of the Daphno sericeae-Cupressetum sempervirentis community, namely Daphne and Berberis ( Figure 6). They are also located on the CCA axis 2 with negative values due to both their low tree cover and the high grazing pressure they suffer ( Figure 5), which leads to the abundance of coprophilous fungi (Sordaria, Spormiella, Cercophora) and anthropozoogenous pollen taxa (Urtica, Plantago lanceolata) in their pollen spectra ( Figure 3).
Probably, the only sample of the Ba2b sub-cluster (CS38) is included within the Ba cluster due to high values of Acer (34%) and Berberis cretica (11%), as well as low percentages of phrygana pollen taxa (Figures 2 and 3). Sub-cluster Ba2b represent upper subhumid mesophilous (PA: 954 mm; TA: 12.2 ºC; Table 1) middle montane mediterranean woodlands (1217 m asl) dominated by Acer sempervirens growing on calcareous rocky sites (limestones), corresponding to the Aceri sempervirentis-Berberidetum creticae association; whose characteristic species are Acer sempervirens, Berberis cretica, Crataegus monogyna and Rosa pulverulenta (Zohary & Orshan, 1965;Barbero & Quézel, 1980;Brullo et al., 2004). According to Brullo et al. (2001), this community can reach the oromediterranean bioclimatic belt close to the treeline (2100 m asl). Probably, the presence of Euphorbia acanthothamnos (2.1%) in sample CS38 would allow us to assign it to the euphorbietosum acanthothamni subassociation described by the previous authors. The high altitude at which this community develops allows certain orophilous chameaphytes such as Prunus prostrata and Rhamnus saxatilis subsp. prunifolia to be frequent. Both pollen taxa are present in the pollen assemblage of this sample (Rhamnus 0.9%; Prunus 1.8%), as well as Crataegus monogyna (3.3%). As attested by Bergmeier (2002), Acer sempervirens is commonly heavily browsed by goats and the spiny Berberis cretica is selectively undergrazed and therefore frequently the dominant species, while the herbaceous layer contains many perennial nitrophytes supported by the dung that is left behind by the sheep seeking the shade of the maple trees. In fact, sample CS38 is also characterized by high values of coprophilous fungi such as Sordaria (8.2%) and Sporormiella (5.7%), as well as by anthropozoogenous pollen taxa such as Urtica dioica (2%) and Plantago lanceolata (2.7%) (Figure 3). The presence of cypress trees in this community is sporadic, hence its low percentage (4%) indicating its regional origin.
Sub-cluster Ba2a is discriminated both in the HCA and CCA (Figures 4 and 5 (Table 7, holotypus, relevé 4). These forests were studied by Brofas et al. (2006) without assigning in any specific syntaxon, but pointing out their differences with other thermo-and meso-mediterranean Cretan cypress forests. The presence of Assulina muscorum in the samples of sub-cluster Ba2b is due to the fact that this type of forest is developed mainly in moist and ombrophilous sites, where this thecamoeba can grow on mosses (López-Sáez et al., 1998). High values of coprophilous fungi and anthropozoogenous pollen taxa are significant of the strong impact of livestock activities on these low altitude forests (López-Sáez & López-Merino, 2007). Samples from Ba2a sub-cluster are arranged close to the relevant pollen indicator taxa ( Figure 6). They are placed close to some samples from Ab (CS47) and Ba1b-1 (CS15 and CS 16) clusters (Pistacio lentisci-sempervirentis) on both CCA axes ( Figure 5) due to their palynological affinities (Figures 2 and 3), mainly because of the significant presence in their pollen spectra of some of the aforementioned meso-thermophilous pollen taxa (Ceratonia siliqua, Myrtus communis, Erica manipuliflora, Olea europaea, Cistus creticus, Rubia peregrina, Smilax aspera, Tamus communis, Hedera helix). The division of cluster Bb provides three main clusters (Figure 4). First, it separates samples with Cupressus/Juniperus values above 60% (cluster Bb1) from others with percentages below 60% (cluster Bb2). Then, the following division of cluster Bb1 discriminates between samples with higher values of Berberis cretica (5−10%) (sub-cluster Bb1a) and those with higher percentages of Quercus ilex/coccifera (> 10%) and phrygana pollen taxa (Sarcopoterium spinosum, Euphorbia acanthothamnos, Thymbra capitata) (subcluster Bb1b). Sub-cluster Bb1a samples represent low to subhumid and orophilous (PA: 875−961 mm; TA: 12.1−13.1 ºC; Table 1) cypress woodlands communities from the lower montane mediterranean bioclimatic belt (1055−1228 m asl) on the northern slopes of the Lefka Ori massif in predominantely E and SE expositions ( Figure 1, Table 1), corresponding to the Luzulo nodulosae-Cupressetum sempervirentis association and its berberidetosum creticae subassociation (Barbero & Quézel, 1980;Bauer & Bergmeier, 2011). This community is the vicariant of the cupressetosum sempervirentis subassociation described above at higher altitude or even at lower altitudes on degraded or deep clay soils. Floristically, the characteristic species of this association are Berberis cretica, Prunus prostrata, Rosa pulverulenta and Rhamnus saxatilis subsp. prunifolia, whose corresponding pollen taxa are documented in their pollen samples ( Figure 2). Interestingly, in some samples of this sub-cluster Bb1a Zelkova pollen is identified, which probably corresponds to the endemic species Zelkova abelicea, one of the most prominent Tertiary relict trees of the Mediterranean region (Kozlowski et al., 2014).
Finally, samples from sub-cluster Bb1b and cluster Bb2 represent xero-thermophilous (PA: 656−997 mm; TA: 11−17.4 ºC; Table 1) cypress woodland communities from the lower supra-to the lower/ middle montane mediterranean belts (529−1525 m asl) of the Lefka Ori massif (Figure 1), exceptionally in the thermo-mediterranean one (284 m asl), growing mainly on limestone and dolomite substrates, sometimes on schists, in dominant S and SE expositions (Table 1). Both group of samples (Bb1b and Bb2) correspond to the aforementioned Luzulo nodulosae-Cupressetum sempervirentis cupressetosum sempervirentis subassociation (sub-cluster Ba1b-3). They are located in different clusters (Figure 4) due to the aforementioned differences in their Cupressus/Juniperus percentage (Figure 2). They are grouped together (also with the samples from Bb1a sub-cluster) along CCA axis 1 with negative values (Figure 5), mainly reflecting their geographical distribution at mid and high altitudes. In short, Ba1b-3 and Bb1b/Bb2 samples correspond to the same community despite being located in very different clusters (Ba and Bb respectively). All of them share floristic affinities, characterized by high constancy in their floristic composition of Quercus coccifera, Acer sempervirens, Luzula nodulosa, Asperula rigida and Pyrus spinosa, maintaining a dense phrygana vegetation cover (Barbero & Quézel, 1980;Bergmeier, 1995;Brofas et al., 2006;Bauer & Bergmeier, 2011). In fact, they all have pollen spectra where phrygana pollen taxa are abundant, mainly Euphorbia acanthothamnos, Phlomis, Sarcopoterium spinosum, Verbascum and Thymbra capitata (Figures 2 and 3). Berberis cretica and Acer sempervirens are also constant species both in the flora and the pollen spectra of samples of the three mentioned clusters, although with lower percentages than in the aforementioned berberidetosum creticae subassociation (sub-cluster Bb1a). As mentioned, the Ba and Bb clusters are differentiated by the percentage of Cupressus/Juniperus pollen, higher than 50% in the second; probably because the presence of Pinus brutia pollen is notable in the two samples of cluster Ba1b-3 coming from the Samaria gorge (Figures 1 and 2).

Conclusions
Our study provides a set of modern surface pollen data spectra which contributes to improving the lack of data from the Aegean islands. Our results clearly demonstrate that it is possible to obtain distinct pollen and non-pollen palynomorph markers for the Cupressus sempervirens forests of the island of Crete. These distinctions are related to specific climatic or geographic conditions, as well as to different degrees of human impact. Classification of modern pollen samples, by means of HCA, indicated the existence of twelve vegetation units, eight of which were ranked as associations and four as subassociations. CCA ordination diagrams of modern pollen samples and pollen/non-pollen palynomorph taxa indicated that palynological differentiation was attributed mainly to factors such as altitude, annual temperature, annual precipitation, tree cover and grazing pressure. On the first CCA axis orophilous tree and shrub pollen taxa (Zelkova, Prunus, Rhamnus, Berberis, Daphne) are separated from low-elevation meso-thermophilous taxa (Myrtus, Olea, Calicotome, Cistus creticus, Pistacia lentiscus, Erica manipuliflora, Prasium, Ceratonia, Laurus, Ruscus, Smilax, Tamus, Hedera, Quercus pubescens and Erica arborea) located at the positive side of the axis, following an altitudinal gradient and increasing precipitation (negative values) or increasing temperature (positive values). On the second CCA axis pollen taxa from tree species (Laurus, Quercus pubescens, Pinus brutia and Quercus ilex/ coccifera) are located on the positive side (increasing tree cover), while most shrubs and herbs are located on the negative side of CCA axis 2.
Our modern pollen rain studies and multivariate analyses allow us to discriminate the peculiarities of six Cupressus sempervirens forest communities on the island of Crete (Greece), proposing three new phytosociological associations: Pistacio lentisci-Cupressetum sempervirentis, Querco ilicis-Cupressetum sempervirentis and Erico arboreae-Cupressetum sempervirentis.
Although Bauer & Bergmeier (2011) does not recognize the two new associations proposed by Brullo et al. (2001) as Daphno sericeae-Cupressetum sempervirentis and Junipero oxycedri-Cupressetum sempervirentis, considering them synonymous with the Luzulo nodulosae-Cupressetum sempervirentis community, our study shows that the associations referred by Brullo et al. (2001) are recognizable by their pollen spectra and are distinguished in both the HCA and the CCA. However, it is true that Daphno sericeae-Cupressetum sempervirentis modern pollen samples show some overlap with those of the Luzulo nodulosae-Cupressetum sempervirentis on both CCA axes. Future studies and the analysis of a higher number of modern pollen samples may contribute to shed light on these potential synonyms.