The inventory started from the most recent and continuously updated list of the vascular plant species and subspecies endemic to Italy (Peruzzi et al. 2014, 2015). The term “endemic” is here used for those taxa that are native to the Italian national territory or to Italy and Corsica (France), as reported in the most recent checklist of the native Italian flora (Bartolucci et al. 2018). All taxa were also checked in Euro+Med PlantBase for name, taxonomic status, and distribution (Euro+Med 2006).

In line with Heinken et al. (2022), we define forest as a “tree-covered areas with normally at least 30% canopy cover (but including, for example, pine forests on sand dunes and bogs, forests on dry rocky slopes, and pasture woods). In the case of completely closed canopies, the area of a stand should be at least a circle with the radius of the maximal height of its tree layer. In the case of an open canopy, the minimum area increases proportionally with the decrease in overhead shading”. The forest habitat includes clear-cuts or temporarily open areas such as forest gaps; stands in the regeneration phase or with shoots from stumps after coppicing; forest edges, including edges of forest tracks. All forest types considered in this work are among those listed in the EUNIS habitat classification system (as “forest and other wooded land”, code T; EEA 2021). Identification of the taxa that are primarily found in forest habitats as defined above was then based on the following sources: 1) Information about the type locality (“locus classicus”) reported in the protologue (original description) of each taxon, based on Peruzzi et al. (2015). The analysis of loci classici is an important source of information to implement strategies of conservation of endemic plants and their habitat (Domina et al. 2012; Brundu et al. 2017). Any reference to forest habitats in the protologue of a given taxon was assumed to be a first and unequivocal indication that the taxon is primarily found in forests; 2) Information from the literature, especially Flora d’Italia Second Edition (Pignatti et al. 2017–2019), relevant books about Italian forests and vegetation (Pignatti 1998; Blasi 2010), papers on single taxa or small groups of taxa, and various web sources; 3) Herbarium collections in FI (Herbarium Centrale Italicum, Firenze), which were used to extract information from over 100 taxa on the type of habitat where collection was done (forest vs non-forest); 4) Hundreds of vegetation plots by the authors, both published and unpublished, which were used to infer occurrence data of the taxa in forest habitats (available from the authors upon request); 5) Personal knowledge by the authors based on years of field work and ad hoc field surveys and observations across the Italian woodlands. Expert assessments are invaluable and are still gaining importance in ecological applications and decision-making in biodiversity conservation (Kuhnert et al. 2010; Heinken et al. 2022).

We included in our list only two categories of taxa: 1) those mainly found in closed-canopy forests, regardless of their type, dynamic stage, structural features, and conservation status, and 2) those occurring in forests as well as in their margins, gaps, and clearings. These two groups largely correspond to the categories 1.1 and 1.2 of the classification recently developed for central and northern Europe by Heinken et al. (2022), and collectively form the group of the so-called “forest specialists”. In addition, all endemic tree taxa found in forests were included in the list (category 1.1), as well as taxa that grow in wet riparian communities of mountain streams and rivulets in shady forest habitats (e.g. Cryptotaenia thomasii and Petagnaea gussonei). In the most recent EUNIS habitat classification system (EEA 2021), these communities are classified in the category “Grasslands and lands dominated by forbs, mosses or lichens” (code R).

For each taxon, information was collected about five major aspects to build a spreadsheet database similar to the EvaPlantE database (Endemic Vascular Plants in Europe; based on Bruchmann 2011): taxonomy, distribution and ecology (habitat), biology, functional traits, and conservation. For each aspect, the following information was recorded.

Taxonomy: 1) existence of a validly designated type specimen (holo-, lecto-, or neotype), mainly after Peruzzi et al. (2015), and later papers; 2) family; 3) order; 4) major clade according to APG IV (for angiosperms) (Angiosperm Phylogeny Group 2016); 5) taxonomic level, as this is relevant when quantifying endemism (Bruchmann 2011): monotypic genus (a single species worldwide, after Mabberley 2017), oligotypic genus (≤ 5 species), species, subspecies, and microspecies. The latter refers to those taxa described and currently accepted at the species level in the national lists above, but belonging to critical taxonomic groups with strong incidence of apomictic reproduction (e.g. Hieracium) and very weakly differentiated on morphological grounds. Moreover, we also indicated the taxa for which doubts about their taxonomic status still exist, based on Bartolucci et al. (2018), the Portal to the Flora of Italy (https://dryades.units.it/floritaly), and the Acta Plantarum project (https://www.actaplantarum.org).

Distribution and ecology: 1) country-level distribution in the Euro-Mediterranean region based on Euro+Med PlantBase; this was done to check for discrepancies with respect to endemic taxa in our list (see above); 2) presence across the 20 Italian administrative regions (shown in Fig. 1 and listed in Table 1), after Bartolucci et al. (2018), the Portal to the Flora of Italy, and Acta Plantarum; and across the five Italian geographic sectors, according to the Nomenclature of territorial units for statistics (NUTS, https://ec.europa.eu/eurostat/web/regions-and-cities/overview), i.e. northwest, northeast, centre, south, and islands. For each administrative region we determined: i) forest endemic species richness (E), as the number of regional (restricted to the region) and national endemics (found in more regions), ii) endemism density in relation to the total regional area (A, km²) and to the forest regional area (F, km², based on Inventario Forestale Nazionale, INFC, https://www.inventarioforestale.org), as ratio E/A and E/F, respectively; the α-index of Hobohm (2003), based on the species-area relationship, was also determined to compare forest endemism density across regions, accounting for their different size; iii) endemism rate as ratio E/S, where S is the regional richness in native species. Next, we used a GLM regression analysis to examine the relationships between regional forest endemic richness (E) and regional area (A, using log function), regional forest area (F, log function), and latitude of the central point of the region (UTM system); 3) altitudinal range based on Pignatti et al. (2017–2019), Acta Plantarum, and personal observations; 4) the dominant or more frequent tree species in the forest type(s) inhabited by each Italian endemic taxon; 5) major reference phytosociological syntaxa (orders or alliances), based on literature, and for the available taxa, Prodromo della Vegetazione Italiana (https://www.prodromo-vegetazione-italia.org); 6) habitat type(s) according to the EUNIS habitat classification system (EEA 2021), using both 2012 and 2021 codes; 7) habitat type(s) according to the EEC Directive 92/43 Annex I; 8) ecological group after Heinken et al. (2022), e.g. whether 1.1 or 1.2 as explained above; 9) type of preferred substrate (indifferent, siliceous, calcareous, basalt), when known also from our field observations; 10) Ellenberg ecological indices (L,T, C, H, R, N), after Pignatti et al. (2005) and Guarino et al. (2012), when available.

Biology: 1) Raunkiaer life-form after Pignatti et al. (2017–2019); 2) chromosome number (2n), when known/available from the Chromosome Counts Database (CCDB version 1.63) (Rice et al. 2015) or Chrobase.it (Bedini et al. 2021). The incidence of polyploidy was estimated by considering as taxa of likely polyploid origin those with chromosome number ≥ 24 divisible by four or more (Coppi et al. 2022), and the data in the updated list of the Italian endemics mentioned above (Peruzzi et al. 2015). We also took into account the presumed ancestral chromosome number of the most recent common ancestor of the family for each angiosperm taxon as reported in Carta et al. (2020); 3) main presumed reproductive system, e.g. whether prevalently gamous or agamous; 4) main pollen vector (wind vs insects); 5) main fruit type; seed dispersal modes were not indicated since documented information about this aspect is at present almost completely missing for the taxa in our list, though partly inferable from the fruit traits and data about related taxa in the same genera.

Functional traits: 1) currently available trait data were checked for each taxon using the TRY database (Kattge et al. 2020).

Conservation: 1) IUCN Red list status according to the IUCN categories and criteria (IUCN 2012), after Orsenigo et al. (2018, 2021).

The list of forest vascular plants endemic to Italy with their respective category is provided in Table 2, while the associated information is given in Supplementary material 1. This list includes 96 species and 38 subspecies (of which 14 doubtful), belonging to 40 families (31 dicots, 8 monocots, 1 gymnosperm). Our results are different from Euro+Med PlantBase for 10 taxa (7.5%), which were missing in the latter, 4 (3%) that were categorized as doubtful, and 9 (6.7%) that were reported as synonyms of other taxa also occurring outside Italy. Taxonomic distribution across higher taxonomic ranks was highly uneven. Asterids, Rosids, monocots, and Ranunculales were, in decreasing order, the most represented major clades, with Asteraceae, Ranunculaceae, and Orchidaceae as most represented families; Boraginaceae, Salicaceae, and Rosaceae were also present in high numbers. As many as seven genera were oligotypic, among which Aegonychon, Limodorum, Cryptotaenia, and Zelkova, and one was monotypic, Petagnaea (Fig. 2). On the other hand, the incidence of microspecies in large genera such as Hieracium, Epipactis, and Ranunculus was also significant (26%).

As many as 81 taxa (60%) were restricted to only one Italian region, while only 3.8% occurred in at least 50% of the regions, and a few others were found over most of the Italian peninsula. The number of endemic taxa per region ranged from 0 (Val d’Aosta) to 44 taxa (Sicily) (Fig. 1). Forest endemism density (E/A) was highest in Calabria and Basilicata and lowest in Lombardy, a large northern region poor in forest endemics (Table 1). Values of the α-index showed considerable variation, being over three times higher in Sicily, Calabria, and Basilicata than in Lombardy (Table 1). Density related to the regional forest surface (E/F) was also highest in Sicily and lowest in Piedmont and Lombardy, also due to the larger forest surface of the latter two regions. The southern and insular regions (Calabria, Sicily, and Basilicata) also had the highest forest endemism rates with respect to the regional species richness (E/S values > 1; Table 1). The rates of forest and non-forest Italian endemism were positively related (p < 0.001; Fig. 3A), though some regions with a similar rate of non-forest endemics showed different rates of forest endemics, for example Lombardy vs Emilia-Romagna. Forest endemism did not depend on regional species richness (p > 0.05) and regional area (p > 0.05) but was weakly related to regional forest area (p = 0.025; Pearson r = 0.49, data not shown). Instead, endemic richness and density increased significantly with decreasing latitude (p < 0.001; Fig. 3B); in fact, the southern and insular sectors hosted both ca 43% of the taxa, the central the 27.4% and the northern ones 18.5 and 9.6 % (northeast and northwest, respectively).

The status of Italian endemics of Alnus cordata and Acer cappadocicum subsp. lobelii needs to be confirmed since the records of these taxa from respectively Albania and former Yugoslavia, which appear in Euro+Med PlantBase, are most likely due to errors (see Barina et al. 2018 and Crowley 2020).

Most of the taxa in our list (58.5%) were linked to closed-canopy forest habitats (category 1.1), while the others were found also in margins, gaps, and clearings (category 1.2). The highest concentration of forest endemics was in the altitudinal range 800–1200 m a.s.l., but a significant proportion also occurred below 800 m (Fig. 4A). About 40 tree species with wide distribution range were found to form the forest habitat of endemics, among which beech was the most important (ca 42.2% of the taxa are associated with this species). Oaks (Quercus cerris, Q. ilex, and Q. pubescens) were other key habitat species (in total 48.9%), and the role of Castanea sativa was also not negligible (8.1%). In terms of syntaxonomical units, the order Fagetalia hosted the largest number of endemics (55.5%; Fig. 5A), especially the southern communities of the alliance Geranio versicoloris-Fagion; the forests of Quercetalia pubescentis-petraeae were also home to a large proportion of endemics (37%), in particular the southern ones of the alliance Pino calabricae-Quercion congestae. Remarkably, hygrophilous forest communities of Populetalia albae, with alliances Salicion albae and Platanion orientalis were found to host a significant number of understorey endemics (14.8%). As many as 34 EUNIS forest types of third or fourth level were habitats for endemics. Of these, “Southern Italian Fagus forests” (T176, 34 taxa, 25%), “Southern medio-European Fagus forests” (T175, 16 taxa, 12%), “Southeastern sub-thermophilous Quercus forests” (T195, 23 taxa, 17%), and “Eastern Quercus pubescens forests” (T193, 14%) were the most important. Hygrophilous forests with Salix (T141, 6.7%) and mesophilous forests with Castanea (T19C, 6.7%) were also relevant habitats.

In relation to the Natura2000 system, forest endemics were recorded in 25 habitats, of which nine are with priority (Fig. 5B). The most important were “Apennine beech forests with Abies alba and beech forests with Abies nebrodensis” (9220*) and “Pannonian-Balkanic, turkey oak-sessile oak forests” (91M0); xerophilous Quercus pubescens forests (91AA*), and “alluvial Alnus-Fraxinus forests” (91E0*) were also significantly represented. Contrastingly, the typical Mediterranean forests dominated by evergreen sclerophylls were comparatively less important habitat for the endemics.

Edaphic preferences have been estimated for ca 64% of the Italian forest endemics. About 40% of these can be found on different soil types, while among the most selective ones, those linked to siliceous soils (acidophilous and/or calcifugous) were significantly prevalent over calciphilous and basophilous taxa mainly growing on limestone or similar calcareous substrates (34.4% vs 11.4%).

Ellenberg indicator values of endemics are still largely incomplete, being available for 44.8% of the taxa in our list. In relation to the light factor, the endemics placed in category 1.1 (closed-canopy forests) were linked to more shaded conditions than those of category 1.2 (also found in margins and gaps), with mean L values of 4.4 and 6.5 (on a 1–12 scale; Pignatti et al. 2005), respectively. As expected, T values were significantly lower in the northern sectors and highest on islands, and vice-versa for C values (Supplementary material 2). Concerning the edaphic factors, southern and insular forest endemics showed a stronger association with acidic and nutrient-poor soils, compared to central and northern endemics.

Hemicryptophytes (ca 44%), geophytes (ca 31%), and phanerophytes (23%) were the most important life-forms (Fig. 4B), while chamaephytes and therophytes were uncommon (1.5 and 0.8, respectively). Among the endemic phanerophytes, nearly half (11.1%) were trees of the upper or intermediate forest layers. This life-form was especially important in the EUNIS habitats T1411 (Mediterranean tall Salix galleries) and T2121A (Southern Italian holm-oak forests). Sicily and Sardinia contain the highest proportion of endemic phanerophytes (51.6%), which represented 44% of the total forest endemism in the insular sector. This form was also important in the southern sector (nearly 20% of the total).

Data about chromosome number are currently available for 51% of the forest endemics, a lower proportion compared to the entire Italian endemic flora (currently ca 63%). Based on these still incomplete data, the range of chromosome numbers was 2n = 10–96 (Paeonia morisii-Symphytum gussonei, respectively), and the frequency of taxa of likely polyploid origin was in the range 40–43% vs 57–60% of diploids. Overall, the diploids/polyploids ratio was considerably higher in the southern (ca 1.5) and even more in the insular sector (2.18), compared to the north-western and central sectors (ca 0.80).

Most understorey endemic taxa are entomophilous, but information about relationships with pollinating insects is currently very scarce. Dry fruits are prevalent, with capsule, achene, and follicle being the most frequent fruit types, followed by the wind-dispersed cypsela of Asteraceae. Fleshy fruits (and false fruits) such as berry, drupe, and pome are found in ca 10% of the taxa, which are typically endozoochorous, as for the taxa in the Rosaceae. Availability of data about functional traits in the TRY database is currently very low. Only 12% of the taxa have been analysed for at least one trait, most often plant height and/or leaf area.

Most of the Italian forest endemics have been assessed for IUCN category (only four taxa not assessed; Fig. 6A), but 23.9% are currently evaluated as DD (data deficient). Most assessed taxa are flagged as LC (least concern) but a significant proportion (31 taxa, 23.7%) is included in the three categories of highest threat, CR (critically endangered, 9 taxa, 6.9%), EN (endangered, 16 taxa, 12.2%), and VU (6 taxa). CR, EN, and VU taxa were mostly assessed based on criteria D (very small populations, < 50 mature individuals), B1 or B2, based on restricted range size. Remarkably, two species of hygrophilous forests of the Po plain in Veneto and Emilia-Romagna, Ranunculus hostiliensis and R. mutinensis, are considered extinct in the wild (category EX). As many as 29 types of threat were found to affect the Italian forest endemics (Fig. 6B); the most important was the increasing frequency and/or intensity of fires (7.1), affecting over 37% of the assessed taxa. “Livestock farming and ranching” (threat 2.3; 25.6%) and “other ecosystem modifications” (7.3; 28.2%) were also recurrent threats.