Local endemisms reflect the portion of biodiversity unique to a particular geographic area and therefore highlights its biological singularity (Good, 1947; Sainz & Moreno, 2002). Although endemic species have concentrated research effort for different scientific interest (Favarger & Siljak-Yakovlev, 1986; Segarra-Moragués & Catalán, 2008; McMullen, 2009) we want to point out two major points. First, their restricted distribution often requires priority conservation politics (Médail & Verlaque, 1997; Dirnböck et al., 2011; Swenson et al., 2012; Huang et al., 2016) and second, they are valuable targets in evolutionary biology (Segarra-Moragués et al., 2007; Ikeda & Setoguchi, 2007; Casazza et al., 2008; Schönswetter & Schneeweiss, 2009; García et al., 2012; Loidi et al., 2015).

Evolutionary processes resulting from geographical isolation or habitat discontinuity may promote the generation of endemisms (Stebbins & Major, 1965; Sunding, 1979). There are different ecological and geographic examples that fit this basic island model in biogeography. Among them, the Northern hemisphere mountain ranges have been intensively studied (Billings & Mooney, 1968; Brochmann et al., 2003; Albach et al., 2006; Winkler et al., 2012), particularly in Europe (Favarger, 1972; Abbot et al., 1995; Schönswetter et al., 2005, Schönswetter et al., 2006). The repeated glaciation cycles during the quaternary have influenced dramatically present vegetation in Europe (Abbot & Brochmann, 2003; Hewitt, 2004; Birks, 2008). Both, transitions from interglacial to glacial state and from glacial to interglacial, increased connectivity between mountain ranges like Sierra Nevada, the Pyrenees, the Alps or Carpathians, which share alpine taxa (Küpfer, 1974; Tutin et al., 1964-1980; Villar et al., 2001 and 2003; Blanca, 2002; Aeschimann et al., 2004; Vargas, 2003; Castroviejo, 1986-2015). These processes have been also documented in the Iberian Peninsula (Loidi et al., 2015; Buira et al., 2017). During interglacial periods, isolation of plant population in high altitudes have been predominant, triggering speciation processes that sometimes were completed before a new glacial cycle started (Rešetnik et al., 2013; Roquet et al., 2013; Deng et al., 2015) and therefore originating endemic species.

Endemic plants in the Pyrenees have been previously investigated by Dupias (1985), Villar & García (1989), Rivas-Martínez et al., (1991) or Villar et al., (1994) and revised in Sainz & Moreno (2002). Recent works in the Alps (Casazza et al., 2008) or the Carpatians (Kliment et al., 2016) developed more detailed floristic and ecological descriptions of the endemisms of these mountains. For a valuable analysis of the Pyrenean endemisms, however, we first require clear criteria to resolve ambiguities on what an endemic taxon is. As Sainz & Moreno (2002) extensively discussed, this is not an easy task, due to the intensive taxonomical revision required, particularly at the level of subspecies or varieties. Local botanists have often concentrated a high effort in the description of local variants, which potentially led to enlarge the list of endemisms. It is illustrative the case of the endemic subspecies Oxytropis campestris subsp tirolinesis described in Schönswetter et al., (2004), for which it was demonstrated no genetic differentiation from the extensively distributed subspecies.

In this paper we present a basic floristic description based on an updated and revised list of endemisms of the Pyrenean flora (Gómez et al., 2016) as a step forward in the description of the biodiversity of the second most important mountain chain of Europe from a biodiversity point of view. To complete this description we also conduct a comparative analysis of the endemisms against the non-endemic flora of the Pyrenees.

Initially we recorded all taxa previously described as endemic of the Pyrenees, including species, subspecies and different apomictic microspecies. Unfortunately, not all proposed endemic taxa have been clearly resolved with unambiguous taxonomic works. For instance, deep molecular research is needed to evaluate the taxonomic entity of all varieties described in apomictic genus and the inclusion of them as endemic taxa might include considerable uncertainty and bias in the floristic and ecological descriptions and analysis here presented. For that reason we discarded microespecies or subspecies from genus such as Achemilla, Hieracium or Pilosella. We also excluded the cases in which only a subspecies of a widely distributed taxon is endemic to the Pyrenees. In this way we attempt to buffer the "overfragmentation" of taxa traditionally linked to taxonomy.

Strictly, a Pyrenean endemism is a species whose distribution area is restricted to the Pyrenees. For that reason, we excluded Pyrenean species of recently discovered punctual localities in mountain ranges far apart from the Pyrenees that have been traditionally considered endemic taxa. That is the case of Veronica aragonensis Stroh, recently cited by Martínez-Ortega (1999) in South Spain (La Sagra, Granada). Nevertheless, we included in our analysis subendemisms, i.e., species whose major distribution area is restricted to the Pyrenees but occur at punctual localities in peripheral areas like Ramonda myconi (L.) Rchb. cited by Calduch et al., (2007) in the Montsiá (Tarragona) or Saponaria caespitosa DC. occurring in the Cantabrian Range.

After screening all Pyrenean taxa, we generated a list of endemisms matching the above mentioned criteria (Gómez et al., 2016). In order to investigate the endemic taxa of the Pyrenees and based on the information recorded in the "Atlas de la Flora Vascular de los Pirineos" (http://www.atlasflorapyrenaea.org) we compiled for all species different biotic and abiotic information such as family, altitude, altitudinal range, vegetation belt, geographic distribution at 3 levels (administrative, biogeographic sector and 10x10 Km UTM grids), abundance, habitat, habitat naturalness, life form and edaphic affinity. We carried out descriptive analysis and comparisons with entire and the non-endemic Pyrenean flora based on Gómez et al., (2017).

Initially, we revised the number of species, genus and families to which the Pyrenean endemisms belong and explored in which administrative regions and biogeographic sectors they are present. The geographic distribution analysis of the Pyrenean endemic taxa was carried in two ways. First we classified each species in one of the 63 possible distribution cases combining 6 sectors. In this way it is possible to explore patterns of distribution within the endemic Pyrenean taxa and to quantify the distribution break events between contiguous sectors. Besides, we plotted the total number of endemisms in each 10x10 Km UTM grid in order to visualize the areas where the endemisms concentrate.

The regional abundance scored for each taxon was tested considering both the whole Pyrenees and only those sectors where the taxon occurs. To test if the distribution of Pyrenean endemisms according to different ecological factors was significantly different from the non-endemic Pyrenean flora we performed Chi-square tests (Table 1).

A careful and integrative revision of the Pyrenean flora reveals that this mountain range does not host a particular high frequency of endemisms, as only 2.4% of native species in the Pyrenees would be restricted to this territory. This number could increase up to 5% when taking into account all microtaxa included in apomictic genus and the endemic subspecies belonging to taxa with a wider distribution range, as already documented by Villar & García (1989). In any case, the number and percentage of endemisms in the Pyrenees contrast with the 477 endemics and subendemics –included apomictic taxa- listed in the Alps (10.8%, Aeschimann et al., 2004), or the 215 listed in the Carpatians (5,5%, Kliment et al., 2016), and it is more similar to the about 90 taxa-including apomictic ones— listed in the Sierra Nevada (4.2%, Blanca, 2002) and higher than the 1-2% reported for the Cantabrian Range (Jiménez-Alfaro, 2008). Nevertheless, further investigations clarifying the taxonomy of the excluded taxa may slightly increase the number of endemic taxa.

Overall, percentages of vascular endemisms in mountain areas are far below of those of oceanic islands such as Canary Islands (25%, Izquierdo et al., 2004), or Madagascar (over 50%, Madagascar Catalogue, 2017), but closer to 7% described in Mediterranean islands like Corsica (Médail & Verlaque, 1997) or Baleares (Govern des Illes Balears, 2012). This pattern suggests that mountain areas are not as isolated to behave as strong refuge as thought. The glacial cycles of the quaternary, which are proposed to promote speciation and radiation of different genus (Koch et al., 1999; Cires & Fernández, 2015), need to be understood as a whole, taking into account glacial and interglacial cycles, which might allow or promote plant migration and therefore reduce the number of narrow distributed mountain endemisms. Buira et al., (2017) reported that Iberian endemisms tend to concentrate in the Northern Western mountains, but with low number of species reduced to a single mountain chain. For instance Villar et al., (1994) recorded 60 vascular plants endemic to the Pyrenees and the Cantabrian mountains, a similar number to the exclusively Pyrenean endemisms. Similar results are described by Jimenez-Alfaro (2008) in the Cantabrian Range. This issue needs further investigation in order to research not only the role of the Pyrenees as a refuge during the glacial maximum, but also as a radiation center in the interglacials (Hewitt, 2004; Loidi et al., 2015).

Radiation processes have been already described in the phylogeny of several genus of families like CARYOPHYLLACEAE (Oxelman et al., 1997; Greenberg & Donoghue, 2001), SAXIFRAGACEA (Deng et al., 2015), PRIMULACEAE (Martins et al., 2003), BRASSICACEAE (Bailey et al., 2006; Rešetnik et al., 2013), which concentrate endemisms in the North hemisphere mountains. In the Pyrenees, endemisms are also concentrated and restricted to few genus and families, included the above mentioned, confirming that radiation processes are not random events, but are evolutionary favored in some lineages.

From an evolutionary perspective, the study of geographical distribution of the endemisms is very informative. It is remarkable that the distribution of the Pyrenean endemisms is particularly well known, as illustrated by the significantly higher number of records per UTM grid they exhibit. Although more attention has been traditionally paid to endemic taxa, our data show a significantly more constricted distribution than the non-endemic flora. Therefore, an ideal complete knowledge of the distribution of all Pyrenean taxa might only accentuate this difference. Another remarkable finding is that all Pyrenean endemisms present continuous distributions at regional scale. The more restricted area that Pyrenean endemisms occupy, however, contrasts with their abundance when present, which is similar to the non-endemic set of plants in the Pyrenees. Moreover, the endemic Pyrenean element presents wide altitudinal amplitude, which is not significantly different from the non-endemics, indicating that the endemic taxa are widely distributed where they are present. These results suggest that the perception of rarity linked to the endemisms is more related to their reduced geographic distribution than to being rare in the environments where they occur.

Our results, therefore, evidence that the distribution areas of the endemisms are locally continuous and geographically restricted within the Pyrenees, but locally abundant as the non endemic flora. These two facts suggest an effective short distance dispersal capacity (sexual and vegetative) together with a strong limitation for long distance seed dispersal of this group of taxa, which may be a clue to understand endemic distribution. Ecological factors like the seed dispersal mechanisms of these endemic taxa would help to understand its aggregated distribution. On the other hand, a clear fragmentation was found in the endemic element between the western and eastern Pyrenees, as for the entire Pyrenean flora (Izard, 1985). In the western Pyrenees the Atlantic and oceanic influence is more dominant, whereas in the eastern Pyrenees the Mediterranean influence is preponderant. The central Pyrenees are indeed characterized by more continental conditions. This longitudinal climatic variation seems to limit the migration of many plant taxa across the East-West axis due to restriction or absence of suitable habitats. Oppositely, latitudinal fragmentation is diminished although the altitudinal barrier might suggest a strong North-South differentiation in the Pyrenees. It is relevant the contrast with the Alps, where the endemisms are predominant in the peripheral areas (Schoswetter et al., 2005) whereas in the Pyrenees they are more abundant in the highest part of the territory.

As described previously for a fraction of the range (García & Gómez, 2007), Pyrenean endemics preferably occupy rocky habitats, and elevations of 1.300-2.500 meters. As previously discussed, it has been largely proposed that mountains promote isolation of the flora in the context of the glacial cycles of the quaternary (Brochmann et al., 2003; Hewitt 2004). This could easily explain why endemisms are rare in lowlands and concentrate above the montane vegetation belt, above the 1.500 meters. These isolation processes are also habitat dependent, and rocky and grassy habitats, where the Pyrenean endemisms concentrate, have been historically among the most patchy and isolated habitats in the mountain regions (Billings, 1968).

One of the endemics, Borderea chouardii, has been intensively studied (García et al., 2002 and 2012) because it is restricted to a single Pyrenean location in the entire world and therefore considered highly endangered. Nevertheless, there are other endemic species very restricted in distribution that might be more sensitive to local perturbations which remain poorly researched. For example, Armeria euscadiensis, Erigeron cabelloi, Polygala vayredae, Primula subpyrenaica or Seseli farrenyi should be investigated to describe their basic biological traits and demographic trends in order to adopt appropriate conservation strategies.

The results of the geographic distribution of endemisms have administrative implications. From a conservation perspective, the highest responsibility lies on large regions like Aragon, Catalunya or Haute-Pyrénées, which concentrate most endemisms. These regions represent most of the area of the Pyrenees. Geographically marginal and smaller regions like Euskadi, Aude and Navarra host less endemic taxa as expected by their location and size.

In summary, after a careful and integrative review of the Pyrenean flora using near 2 millions of records, the endemism rate turned out to be lower than expected and published so far, although not so different to other European mountain floras. Our results about geographical distribution patterns suggest that isolation during the glacial cycles might have not been as strong as expected either. Pyrenean endemic plants tend to be chamaephytes which concentrate in grasslands and rocky habitats, on limestone substrates at 1.200-2.500 meters of altitudes. There is a clear taxonomic bias in families and genus represented in the Pyrenean endemisms towards some alpine groups, which are also more sensitive to speciation processes in other mountain chains of the Norther Hemisphere. Further and deeper analysis, integrated with neighbour mountain chains might conduct to new understanding of the evolutionary process which led to the present flora distribution of the Pyrenees. Finally, we consider that the accurate knowledge of the Pyrenean endemic taxa distribution should help to adopt effective conservation strategies.