Woods with Quercus petraea (Matt.) Liebl. in Tuscany (Italy): a vegetation classification approach

. In Tuscany (Italy), oak mixed woods with high cover values of Quercus petraea are rather infrequent but well documented within local and national phytosociological studies, even if, in the literature, not always analyzed and well characterized from the syntaxonomic view point. We gathered 71 published and unpublished Tuscan phytosociological relevés where Q. petraea was dominant or with relevant cover values, that were investigated by means of multivariate analysis. The ecological requirements of the resulting groups were indirectly calculated by means of Ellenberg Indicator Values (EIV), and the fidelity coefficient (PHI) for the diagnostic species of each group was calculated. According to our analysis, five different types of Q. petraea woods were found to be present in Tuscany. Each group was characterized floristically and ecologically, allowing to investigate their syntaxonomic aspects. Thus we have attributed the Tuscan Q. petraea communities to five different associations, two of which already existing and three are here described as new associations. In the end, some conservation aspects of these woods regarding Natura 2000 habitats are discussed. Recopilamos 71 datos fitosociológicos de esta región publicados y no publicados en los que Q. petraea aparece como dominante o con valores de cobertura relevantes y que se han analizado mediante análisis multivariante. Los requisitos ecológicos de los grupos resultantes se calcularon indirectamente mediante los valores del indicador de Ellenberg (EIV), y se calculó el coeficiente de fidelidad (PHI) para las especies de diagnóstico de cada grupo. Los resultados del análisis mostraron cinco tipos diferentes de bosques de Q. petraea en la Toscana. Cada grupo está caracterizado por un elenco florístico concreto y unos requisitos ecológicos diferenciados que nos ha permitido investigar sus aspectos sintaxonómicos. Así podemos atribuir las comunidades toscanas de Q. petraea a cinco asociaciones diferentes, dos de las cuales ya existen y tres son descritas en este trabajo como nuevas asociaciones. También se discuten algunos aspectos de conservación de estos bosques con respecto a los hábitats Natura 2000.


Introduction
Quercus petraea (Matt.) Liebl., which encompasses a group of several infraspecific taxa, has a fundamentally European distribution, with some ramifications towards southern Europe and the Mediterranean basin (www. emplantbase.org). In Italy, it can be considered as a rather common species in the Alps and pre-Alps but it becomes more and more sporadic proceeding southwards (Andreis & Cerabolini, 1993;Viciani & Moggi, 1997;Andreis & Sartori, 2011;Viciani & al., 2016a). In Italy Q. petraea generally lives in forests dominated by other tree species, and the woods where it can be found to have a relevant cover value or to be the dominant tree are very rare, especially in central and southern part of the country. The sociological aspects of Italian forests in which this oak plays an important role have been subjected to several past studies, but mainly based on local approaches that have led to an extremely differentiated and fragmented syntaxonomical framework (e.g. Oberdorfer & Hofmann, 1967;Pedrotti & al., 1982;Brullo, 1984;Scoppola & al., 1990;Arrigoni, 1997;Viciani & Moggi, 1997;Biondi & al., 2002;Di Pietro & al., 2010;Andreis & Sartori, 2011). In a recent study, Viciani & al. (2016a) analyzed the coenological and chorological features of the Italian communities dominated by Q. petraea with respect to the European context, and highlighted the presence of homogeneous floristic, chorological and ecological groups that, from the syntaxonomical point of view, allowed to attribute the Italian Q. petraea woods to seven alliances of three different orders. From this starting point, we focused on Tuscan relevés and investigated the floristic-ecological traits of Q. petraea woods, with the aim to search for the presence of floristic-sociological different groups, identify the diagnostic species of these groups and propose a classification at the association level for the communities of this territory.

Data collection and analyses
The study followed the standard approach of the Zürich-Montpellier School (Braun-Blanquet, 1964;Westhoff & van der Maarel, 1978) according to fundamental and updated concepts recommended by Dengler & al. (2005), Dengler & al. (2008), Biondi (2011), Pott (2011. We searched in all the Tuscan vegetation literature for phytosociological relevés in which Q. petraea was present and we added unpublished relevés owned by us. The first dataset consisted of more than 100 relevés. Among all these relevés we then selected only those in which Q. petraea played an important role in the canopy, i.e. where Q. petraea had cover values 3, 4 and 5 of Braun-Blanquet scale, in order to avoid doubtful environmental correlations with low cover of this oak. After this procedure, the resulting subset was composed by 71 relevés (of which, 19 were unpublished), which constituted the dataset for the analyses. The distribution of the sampled sites is graphically shown in Figure 1, while more precise locations and data of the relevés, together with bibliographical references, are reported in Supplement 1. The sporadic species, i.e. species present in less than four relevés (< 5%) and with cover-abundance values ≤ 2 were disregarded in the numerical analyses. The Braun-Blanquet cover-abundance scale were transformed according to the ordinal scale proposed by Van der Maarel (1979) and Noest & al. (1989); the resulting matrix (71 relevés × 101 species) was numerically classified by cluster analysis using standard statistical software as SYN-TAX 2000(Podani, 2001 and PAST (Hammer & al., 2001). We performed several algorithms and similarity measures, which gave very similar results. The dendrogram showed was performed with "paired group" as algorithm and Bray-Curtis index as similarity measure.
The diagnostic species among the groups resulting from cluster analysis were statistically defined by the PHI coefficient of association (Chytrý & al., 2002), and performed using the indicspecies package (De Cáceres & Legendre, 2009) in the R environment (R Core Team, 2015). The significance of the fidelity coefficient was verified according to a Fisher's exact test. We considered a species as diagnostic of each group if phi ≥ 0.40 and p < 0.05; the threshold values (phi = 0.40) is chosen because it produces neither too long nor too short lists of diagnostic species for each vegetation unit (Illyés & al., 2007). The number of relevés of each cluster was virtually standardized to an equal size (Tichý & Chytrý, 2006) in order to eliminate dependency of the PHI coefficient for presence/absence data on the relative size of groups within data set.
Exploratory analysis of floristic variations within Tuscan Q. petraea woods as function of environmental variables was performed through Ellenberg Indicators Values (EIV, see Ellenberg & al., 2001;Pignatti, 2005). In particular, we considered the following factors: Light (L), Temperature (T), Moisture (U), Nitrogen (N), Continentality (C) and Soil pH (R). For each relevé, each total indicator value was calculated using the weighted averages of the presence/absence data of the species recorded in the plot (except for sporadic species). Differences in mean EIVs among relevés and groups were displayed by NMDS analysis, with Simpson index as similarity measure, performed using PAST statistical software (Hammer & al., 2001).

Results and Discussion
The Tuscan Q. petraea wood relevés are reported in Supplement 2, grouped according to the results of cluster analysis (dendrogram of Supplement 3). We can distinguish five principal groups within the data set. The groups are well separated on a floristic basis, as showed by the results of PHI analysis (Table 1), and also distinguished on ecological bases, as highlighted by the NMDS analysis ( Figure 2). The length and the orientation of the vectors associated with environmental variables depicted their importance in structuring Q. petraea vegetation types in Tuscany ( Figure 2). The vector associated with continentality (C) resulted to be very short, thus suggesting that this variable had overall low importance in structuring woodland vegetation, while the most important ecological factors appeared to be light (L), soil pH (R) and nitrogen availability (N) (Figure 2).
The first main division of the dendrogram (Supplement 3) separates the relevés in which presence and cover values of the Mediterranean species (Quercetea ilicis) are less relevant from the relevés in which these components are better represented (Supplement 2). In the first subcluster, two main groups can be separated. The first one (Group 1) encompasses rather mesophilous woods, while the second one (Group 2) comprises relatively open woods on very acidic and nutrient-poor substrates ( Figure 2). The second main division of the dendrogram encompasses relatively more thermophilous and less acidic woods that can be separated in three subclusters (Groups 3-5), always on floristic and ecological bases (Table 1, Figure 2).
We prepared two synoptic tables, one to compare the Tuscan groups of relevés each other (Supplement 4), and another comparing the detected Tuscan groups with the most similar associations from literature (Supplement 5).   Syntaxonomy: relevés very similar to those of Group 1, located in a sub-Mediterranean hilly environment in Emilia Romagna, were described by Ubaldi & al. (1995) as a new association named Serratulo tinctoriae-Quercetum petraeae. We can note some floristic minor differences (i.e. a greater cover of Quercus pubescens and the presence of Iris graminea, Symphytum tuberosum, Viburnum lantana, and others in Emilia conenoses, see Supplement 5), but the specific-characteristic composition is very close to that one of Group 1, so we can use this syntaxon. A nomenclatural issue must be solved, because in a later work, Ubaldi (2003) considered Serratulo-Quercetum a new association, based on the same table of Ubaldi & al. (1995) and separated the original table in two subassociations, typicum and paeoniaetosum, providing a new type relevé also for the subassociation typicum. This is in contrast with the International Code of Phytosociological Nomenclature (Weber & al., 2000), because the Serratulo-Quercetum typicum resulted to be correctly described and typified in Ubaldi & al. (1995).

Group 2.
Mixed and relatively open woods of Q. petraea, Q. cerris and Pinus pinaster, occurring in central and central-southern Tuscany (Figure 3), on acid and poor-nutrient soils (Figure 2), mainly coming from "Verrucano" geological unit (metaconglomerates, metasandstones, quartzites, metavolcanic rocks, etc.). Many acidophilous shrubs and herbs occur in the dominated layers, together with some mesohygrophilous species (Supplement 2). Two different aspects, influenced by human management, can be identified: a first one, with P. pinaster absent or sporadic (subgroup 2a, mainly located in Merse and Farma valleys, central-southern Tuscany); a second one, having a more open canopy, characterised by a stronger presence of P. pinaster in the dominant layer and various shrubs (especially Rubus sp.) in the undergrowth, due to frequent coppicing and conifer plantations (subgroup 2b, mainly located in Cerbaie site, central Tuscany). The last four relevés on the right of subgroup 2b are somehow differentiated, as results from the dendrogram, probably due to the presence of Quercus robur and other mesohygrophilous species.  Syntaxonomy: these communities can be attributed to the association Frangulo alni-Quercetum petraeae Arrigoni in Foggi & al. 2000described by Arrigoni (1997 for Cerbaie, Tuscany (with the name Ilici aquifolii-Quercetum petraeae Arrigoni 1997 nom. illeg.). In our analysis this vegetation type completely overlaps with Hieracio racemosi-Quercetum petraeae fraxinetosum orni Arrigoni 1997, described for the same area by the same author: the type relevés of Frangulo alni-Quercetum and Hieracio racemosi-Quercetum fraxinetosum fall within the subgroup 2b (rel. n. 59 and 67 in Supplement 2). Frangulo-Quercetum is rather similar to Hieracio racemosi-Quercetum petraeae Pedrotti, Ballelli & Biondi 1982 (described for Gubbio basin, Umbria) from which it differs for the presence of several thermophilous species lacking in Gubbio relevés (see Supplement 5). Ubaldi (2003) noted this difference and using Arrigoni's relevés of Hieracio racemosi-Quercetum described a new association named Lonicero etruscae-Quercetum petraeae Ubaldi 2003, which in our opinion must be considered a synonym of Frangulo alni-Quercetum petraeae.

Group 5.
Relatively open Q. petraea-dominated woods with abundant Q. ilex, located in central Tuscany (Figure 3) on soils derived by ultramafic substrates, therefore not so acidic (but with Mg instead of Ca) and nutrient-poor ( Figure 2). The dominated wood layer is rich in sclerophyllous dry-tolerant species and scarce in Q. cerris, which on ophiolite rocks seems to be less competitive than Q. petraea (Supplement 2).
Syntaxonomy: these peculiar communities were not described before and syntaxa with similar characteristics are not present in literature, so we propose to attribute them to a new association named Junipero oxycedri-Quercetum petraeae ass. nova hoc loco, holotypus rel. 3 in Table 4 (rel. 67 in Supplement 2).  Mucina & al., 2016), not occurring in Tuscany (Viciani & al., 2016a). Physospermum cornubiense is widely present also in our relevés but, as showed by our analysis, it is not discriminating of any Q. petraea vegetation type in Tuscany. Some communities similar to Physospermo-Quercetum petraeae could have been located in the past also in the Tuscan slopes of the Apennines, but as can be derived from the general distribution of the relevés (Figures 1, 3), at present Q. petraea is rather rare in the Tuscan Apennines and generally does not form woods in which it has relevant covers, because it has been extensively substituted by Castanea sativa plantations at higher altitudes and by agricultural land uses at lower altitudes (Bernetti, 1987;Arrigoni & Viciani, 2001;Viciani & al., 2016a). Due to these considerations and to the above mentioned floristic and coenological features of Tuscan Q. petraea coenoses, in accordance with Viciani & al. (2016a), we therefore propose to attribute all the Tuscan Q. petraea associations to Crataego laevigati-Quercion cerridis Arrigoni 1997 alliance. Probably, Serratulo-Quercetum petraeae, especially in the form of northern Tuscany (Mugello) conenoses, can be considered close to a sort of transition towards Physospermo-Quercion petraeae associations, while Frangulo alni-Quercetum petraeae is not so far from some aspects of the northern Italian Quercion roboris (see Andreis & Sartori, 2011). The attribution of all the Tuscan associations to Crataego laevigati-Quercion cerridis is reinforced also by bioclimatic and phytogeographic considerations, since relevés are all located in a generally sub-Mediterranean environmental context (see Blasi, 2010;. Respect to higher syntaxonomic ranks, the syntaxonomic order generally accepted is Quercetalia pubescenti-petraeae, while at the class level there are some discrepancies between the Italian schemes Biondi & Blasi, 2015) and the European ones (Mucina & al., 2016) that cannot be here solved.

Conservation aspects
The study of vegetation with the phytosociological approach is crucial for detecting many habitats of conservation interest in accordance with the European Habitat Directive 92/43/EEC European Commission, 2013;Viciani & al., 2014Viciani & al., , 2016bAngiolini & al., 2017). This is true also for woods, and the conservation importance of several forest types have been recognized and listed in the Habitat Directive as different types of conservation interest habitats. As to woods where Q. petraea has an important role, the Habitat Directive lists several habitats with different Natura 2000 codes (e.g. for Italy: 9110, 9170, 91L0, 91M0, see Biondi & Blasi, 2009;European Commission, 2013;Janssen & al., 2016). EU habitat types are mostly described at the level of a syntaxonomic alliance (Rodwell & al., 2002;Evans, 2006Evans, , 2010, so all the Tuscan coenoses can be probably mostly attributed to the habitat Natura 2000 code 91M0, named "Pannonian-Balkanic Turkey oak-sessile oak forests" but found to be present also in the Italian peninsula, especially in the Thyrrenian side (see the Italian Interpretation Manual of the 91/43/EEC Directive Habitats -vnr.unipg.it/habitat; Biondi & Blasi, 2009