Xeric grasslands of the inner-alpine dry valleys of Austria – new insights into syntaxonomy, diversity and ecology

Aims: We studied the syntaxonomic position, biodiversity, ecological features, nature conservation value and current status of dry grasslands investigated by Josias Braun-Blanquet more than 60 years ago. Study area: Inner-alpine valleys of Austria. Methods: We sampled 67 plots of 10 m2, following the standardized EDGG methodology. We subjected our plots to an unsupervised classification with the modified TWINSPAN algorithm and interpreted the branches of the dendrogram syntaxonomically. Biodiversity, structural and ecological characteristics of the resulting vegetation units at association and order level were compared by ANOVAs. Results: All the examined grasslands belong to the class Festuco-Brometea. From ten distinguished clusters, we could assign four clusters to validly published associations, while the remaining six clusters were named tentatively. We classified them into three orders: Stipo-Festucetalia pallentis (Armerio elongatae-Potentilletum arenariae, Phleo phleoidis-Pulsatilletum nigricantis, Medicago minima-Melica ciliata community, Koelerio pyramidatae-Teucrietum montani), Festucetalia valesiacae (Sempervivum tectorum-Festuca valesiaca community); Brachypodietalia pinnati (Astragalo onobrychidis-Brometum erecti, Agrostis capillaris-Avenula adsurgens community, Anthericum ramosum-Brachypodium pinnatum community, Ranunculus bulbosus-Festuca rubra community, Carduus defloratus-Brachypodium pinnatum community). Conclusions: The ten distinguished dry grassland communities of the Austrian inner-alpine valleys differ in their ecological affinities as well as their vascular plant, bryophyte and lichen diversity. We point out their high nature conservation importance, as each of them presents a unique habitat of high value. Copyright Martin Magnes et al. This is an open access article distributed under the terms of the Creative Commons Attribution License (CC-BY 4.0), which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited. Vegetation Classification and Survey 2: 133–157 doi: 10.3897/VCS/2021/68594 RESEARCH PAPER International Association for Vegetation Science (IAVS) CLASSIFICATION OF OPEN HABITATS IN THE PALAEARCTIC Martin Magnes et al.: Grasslands of Austrian inner-alpine dry valleys 134 Taxonomic reference: Names of vascular plants, bryophytes and lichens follow Fischer et al. (2008), Frahm and Frey (2004) and Nimis et al. (2018), respectively. Syntaxonomic reference: Names of orders and classes follow Mucina et al. (2016), references for associations and alliances are given in the text. Abbreviations: ANOVA = analysis of variance; DCA: detrended correspondence analyses; EDGG: Eurasian Dry Grassland Group; EIV: ecological indicator value; FL: Fließ; GR: Griffen; GU: Gulsen; KA: Kaunerberg; LA: Laudegg castle in Ladis; MA: Marin; NM: Neumarkt in der Steiermark; OM: Obermauern; PÖ: Pöls; PU: Puxer Loch; TWINSPAN = Two-way indicator species analysis; ZS: Zinizachspitze.


RESEARCH PAPER
International Association for Vegetation Science (IAVS)

CLASSIFICATION OF OPEN HABITATS IN THE PALAEARCTIC Introduction
The European Alps ("Alps") form a nearly 1,200 km long arc from Nice in the southwest to Vienna in the northeast, with the widest sector in the centre of the Eastern Alps, stretching approximately 240 km from north to south (Ozenda 1988). The Alps represent a biodiversity hotspot in Central Europe, with 4,491 observed vascular plant species (Aeschimann et al. 2004; Barthlott et al. 2005). The inner-alpine valleys are the central parts of the great longitudinal valleys between the highest ridges of the Alps and represent islands of continentality within the relatively oceanic climate of the mountain ranges. One important aspect of continentality is the low precipitation, caused by the high altitudes of the mountains that cast rain shadows towards the south and southeast, leeward of the prevailing rain-bringing northwestern winds. While even the driest parts of the Alps are more humid than true steppes (Ozenda 1988), these valleys provide refugial habitats to many species with a distribution centred in the steppe biome of Eastern Europe . Similar intra-montane valleys with continental climate and Festuco-Brometea grasslands are known from other high mountain ranges, like the Caucasus (Belonovskaya et al. 2016).
The inner valleys of the Alps have been inhabited by humans since the Bronze Age (Sydow 1995;Bätzing 2005) or even earlier (Patzelt 1996), up to elevations exceeding 2,000 m (Braun-Blanquet 1961; Schwabe and Kratochwil 2012). Therefore, the contemporary landscape has been influenced by millennia of land use, mainly as pastures and meadows. It is evident that the traditional agro-pastoral practice exerted a positive feedback on the Festuco-Brometea grasslands, as already noticed by Braun-Blanquet (1961), who described the succession from the Astragalo-Brometum to Larix decidua or Pinus sylvestris forests after the cessation of management. Both the isolated relic character of these steppe elements and the requirement of adequate management underline the need of skilled nature protection measures in these unique habitats. Phytosociological studies are well suited for evidence-based decisions in nature conservation. However, most studies of the inner-alpine dry valleys (Buschardt 1979;Wagner 1979;Kratochwil 2004, 2012;Wiesner et al. 2015;Dengler et al. 2019) focused on the more (south)western parts of the Alps with a severe summer drought, and merely touched the Eastern Alps, which have only a moderately continental climate that is influenced by cyclones from the Adriatic Sea.
Our study is part of a long-term project of the Eurasian Dry Grassland Group (EDGG) that will cover the inner-alpine dry grasslands in the whole Alps (Dengler et al. 2020a;Magnes et al. 2020). In the present paper, we focus on the dry grassland vegetation of the inner-alpine valleys of the Austrian Alps, revisiting sites that were already described by Braun-Blanquet (1961). More specifically, we provide a preliminary syntaxonomic classification using modern multivariate methods, describe the biodiversity and ecology of these grasslands and discuss their conservation management.
The sampling was performed in the Nature Park Kaunergrat, southwest of Landeck, in one of the driest regions of the Austrian Alps. Our study area is close to the Engadin in Switzerland and the most continental area of the Eastern Alps, the Vinschgau in South Tyrol (Italy). Braun-Blanquet (1961) described the dry grasslands of the Austrian Upper Inn Valley in conjunction with those of the Lower Engadin because of the proximity and the similarity in geology, climate and species composition.
Geologically, the Kaunerberg (KA) and the study sites below the castle Laudegg (LA) belong to the easternmost part of the Engadin window. Kauns-Kaunerberg is located above Bündner schist, a metamorphic marine sediment from the Cretaceous period that is rich in calcium carbonate (Gruber et al. 2010). In contrast, the study sites below the Laudegg castle are situated above a much older, Triassic bedrock of metamorphic quartzite sandstone (Gruber et al. 2010). However, both rock materials favoured the formation of shallow soils at these localities. Another process that had a large influence on the composition of the soils at the studied sites was the deposition of drift and moraine material by the Inn glacier during the Würm Glacial. In course of this, rocks from a wider glacial catchment (e.g. Bündner schists, but also dolomites) were distributed and deposited along the valley slopes. This propagated the  formation of calcareous soils at the study sites below Fließ (FL), although the bedrock below is exclusively made up of acidic mica schists of the Variscan Silvretta crystalline (Gruber et al. 2010).
The studied dry grassland sites bear witness of an old cultural landscape. Traditionally, these sites were communal pastures ("Allmende") for small livestock (mainly sheep, to a lesser amount goats) in spring and autumn, before and after summering on alpine pastures at higher elevations. This kind of management is documented back to the medieval ages, but likely is much older. Until the 1940s and early 1950s, several hundred animals were herded in the slopes below Fließ. However, this practice was abandoned alongside the economic upturn of the post-war era, leading to a reforestation of the sites. This negative trend lasted until 2002, when, following the long-lasting pressure of regional conservationists, the sites at Fließ were finally put under protection and integrated into the Natura 2000 network. The sites at Kauns-Kaunerberg followed shortly after and are protected under Tyrolean law since 2006. Since then, both sites have been managed by the Kaunergrat Nature Park. Management measures include the initial re-establishment of abandoned areas by removal of bushes and trees, and subsequent, targeted grazing by cattle, goats and sheep. Since their implementation, these activities have been regularly evaluated via monitoring studies on vegetation and selected insect taxa (butterflies, ants). The area is the best example of inner-alpine dry valleys in Austria concerning size and extent of dry and semi-dry grassland vegetation.
The Virgen valley runs in a west-east direction south of the main ridge of the Alps. It is quite remote from modern traffic routes and accessible only from the east, following the Isel river upstream. Due to its position south of the giant mountain massif Großvenediger (3,662 m), the climate is rather dry with an average annual precipitation of about 900 mm. The nearest climate station (Lienz, 30 km to the southeast) shows sub-Mediterranean characteristics with a relatively low precipitation in late summer (Figure 2), while the precipitation in Virgen is probably significantly lower (Wagner 1979). The main settlements are found from 1,200 to 1,300 m. North of the valley the geological bedrock is dominated by the schist shell of the Upper Tauern Penninicum.
The valley has been populated since prehistoric times. Triticum aestivum was cultivated until the 1980s and up to an elevation of approximately 1,400 m. Cattle breeding is another important land use. Traditionally, the animals were brought across mountain passes to markets on the northern side of the main ridge (Oberwalder 1999). Since the 19 th century, markets that are more distant have become relevant. The settlement structure in the valley is characterised by small villages surrounded by a structurally rich cultural landscape called "Virger Feldflur" (Michor 1999). Part of it are hedges with Berberis vulgaris, Rosa spp. and Sorbus aucuparia. The former arable fields were gradually transformed into grasslands, resulting in a saying in local dialect: "Virgn isch grin woagn" ("Virgen has become green") (Oberwalder 1999).
The montane and subalpine zones reach from the valley bottom up to more than 2,000 m. On the northern slopes, Picea abies forests are dominant. On the sunny southern slopes, the forests were replaced by grasslands centuries ago. The forest remnants are usually dominated by Larix decidua with low cover and are still used as reserve pastures for dry years. The often very steep slopes were not only grazed by cattle, sheep and goats, but also used as mountain hay meadows as farmers could not produce enough hay in the lower parts of the valley. Hay was used as forage in winter in addition to harvested leaves and straw. The mountain meadows, sometimes even above the timberline, were traditionally mown only every second year. They are situated in the neighbourhood of the mountain pastures, which are mainly used for young cattle (Oberwalder 1999).
The bedrock of the sampling localities is mainly calcareous slate, partly covered with moraine material, and only on the southern border of OM there is also some dark phyllite together with the calcareous slate (Frank et al. 1987).
The studied part of the Mur valley is situated along the upper reaches of the river, approximately 80 to 120 km east of its source. Braun-Blanquet (1961) identified this area as the easternmost extension of the inner-alpine dry valleys. In comparison to the climate of Lienz (Figure 2), the precipitation is lower without a depression in late summer ( Figure 2, climate station of Zeltweg). Additional to the two sites within the Mur valley itself (GU, PU), we include here plots in the area of an adjacent pass traversal (NM), one site in the Carinthian Basin (GR) and one site in a tributary valley of the Mur with the last population of the endemic Stipa styriaca (PÖ). The latter locality was not investigated by Braun-Blanquet (1961), because this species, probably a remnant of the glacial steppe-tundra, was discovered and described from this place only later (Melzer 1962;Martinovský 1970).
Although all sites are situated in the Austrian Central Alps, the geological bedrock is very diverse (Flügel and Neubauer 1984). While the westernmost site (PU) is located on a steep slope of Palaeozoic limestone (Murauer Palaeozoikum), the bedrock in NM consists of moraine material with quartzite. The habitat of Stipa styriaca in PÖ is a steep southern slope on black mica schist, and GU is part of the greatest connected serpentinite outcrop of Central Europe (Brooks 1987), its steep southern slopes being one of the most peculiar habitats in Austria (Magnes 2018). The castle hill of Griffen (GR), a landmark visible for miles around, is an isolated block of marble (Gleirscher and Pacher 2005).

Vegetation and environmental data
In total, we sampled 67 plots of 10 m 2 (15 nested-plot series with two 10-m 2 subplots each and additional 37 normal plots, Suppl. materials 1, 2) with the EDGG methodology ) between the 6 th and 13 th July 2018. Each nested-plot series ("EDGG Biodiversity Plots") consists of a 100-m 2 plot, with nested series of 0.0001, 0.001, 0.01, 0.1, 1 and 10 m 2 in two opposite corners. All terricolous vascular plants, bryophytes and lichens were recorded with the shoot presence method (Dengler 2008) for all plot sizes, and their cover value in percent was estimated in the 10-m 2 plots.
Soil of the uppermost 10 cm was collected in each plot by mixing five random samples. The following soil parameters were measured (all measured environmental variables as header data are provided in Suppl. material 1): skeleton content (mass fraction of particles > 2 mm), percentages of sand, clay and silt, texture class estimated with Robinson pipette method after removing organic matter with 6% H 2 O 2 , field capacity (%), drainage rate (cm/h), available water (%), saturation (%), permanent wilting point (%), pH (in a suspension of 10 g dry soil in 25 ml aqua dest.), electrical conductivity (EC) (in a suspension of 10 g dry soil in 50 ml aqua dest., dS/m), organic matter (as loss at ignition at 430°C, %), organic carbon (%), N content estimated with the Kjeldahl method (%), C/N, available P (estimated with the Mehlich 3 method (PM3), mg/kg).
The complete data are stored in and available from the GrassPlot database Biurrun et al. 2019; https://edgg.org/databases/GrassPlot) as data-set AT_E. The 10-m² plots are also provided in Suppl. material 1 (header data) and Suppl. material 2 (sorted relevé table).

Data analysis
The compositional data, along with the header data, were entered into TURBOVEG (Hennekens and Schamineé 2001) and then imported into the program JUICE (Tichý 2002). Prior to the numerical analysis, occurrences of species in the shrub and herb layer were combined, using the formula of Fischer (2015). We conducted a TWINSPAN (Two-Way Indicator Species Analysis ;Hill 1979) in the modified version proposed by Roleček et al. (2009), with the following settings: three pseudospecies cut levels (0%, 5% and 25%); total inertia as measure of heterogeneity; minimum group size: 2.
We tried cluster numbers up to 15 and finally selected the solution with 11 clusters as the basis for our classification, as it corresponded most closely to what we consider the association level. One cluster containing a single relevé was joined with another cluster, so finally ten relevé groups were distinguished. Moreover, three clusters were subdivided into regional variants. These variants partly corresponded to TWINSPAN clusters of finer resolution, but the assignment of relevés to regional variants was based on their geographical distribution.
Fidelity of species to relevé groups was calculated using the phi coefficient of association (Chytrý et al. 2002), assuming equal cluster size (Tichý and Chytrý 2006). To provide diagnostic species both at order and association level, the calculations were done in Excel, which precluded the application of Fisher's exact test. However, we set phivalue thresholds so high that essentially no non-significant species were chosen at the association level. Species were considered as diagnostic with phi ≥ 0.25 at order level and phi ≥ 0.45 at association level. Additionally, we required in both cases that the drop in phi-value to the syntaxon with the next highest phi-value was at least 0.25; otherwise, the species was also considered diagnostic in the latter syntaxon. If such a drop of 0.25 did not occur anywhere in the sequence of syntaxa ordered by decreasing phi-values, a species was considered not diagnostic. If a species was diagnostic at both order and association level, it was assigned to the level where the phi-value was higher, provided all other conditions were fulfilled. The term "diagnostic species" is used here in a data-set specific context and should not be understood as equivalent to character species (see Willner et al. 2009).
We tried to assign our relevé groups to phytosociological associations, alliances and orders described in the literature by comparing the species composition. If such an assignment was ambiguous, we used informal community names. We refrained from formally describing new syntaxa, as this will be the task of a future, more comprehensive revision of all dry grasslands of the Eastern Alps. To facilitate the syntaxonomic interpretation, we calculated for each relevé the total percentage cover of the diagnostic species of orders according to Willner et al. (2019) using the function "total cover of <colour> species" in JUICE.
To visualize the relation of the species composition of the relevé groups (and relevés) to measured or calculated factors we performed a DCA using Canoco 5 (ter Braak and Šmilauer 2012) with downweighting of rare species and log-transformation of cover values.
The number and cover of red list species were calculated in JUICE based on Niklfeld and Schratt-Ehrendorfer (1999) for vascular plants, Grims and Köckinger (1999) and Saukel and Köckinger (1999) for bryophytes and Türk and Hafellner (1999) for lichens.

Numerical classification
The first TWINSPAN division resulted in two groups roughly corresponding to the orders Stipo-Festucetalia pallentis (clusters 1-7) and Brachypodietalia pinnati (clusters 8-11) of the class Festuco-Brometea ( Figure 3). Cluster 4 was transitional between the two orders, with most of its relevés having a higher total cover of Brachypodietalia species (Figure 4). Cluster 7 was transitional between the Stipo-Festucetalia pallentis and Festucetalia valesiacae with several relevés clearly belonging to the latter order ( Figure 4). Cluster   Table 1 in the position of cluster 4. Note that the percentage cover of non-diagnostic species is not shown. 6, containing a single relevé, showed a clear prevalence of Brachypodietalia species (Figure 4). As it shared several otherwise rare species with cluster 8 (e.g., Allium carinatum, Avenula adsurgens, Jasione montana, Potentilla argentea), and the relevés of both clusters were from the same location, we joined these two clusters into a single community. Otherwise, we refrained from moving single relevés between clusters based on their affinity to phytosociological orders, but potentially misclassified relevés are mentioned in the description of the individual communities.
Therefore, we classified our relevés into ten vegetation units at association level, which in turn were grouped in the three orders (A) Stipo-Festucetalia pallentis, (B) Festucetalia valesiacae and (C) Brachypodietalia pinnati (Table 1, Suppl. material 2). The alliance level remained unresolved (see Discussion).

Ordination
The first axis of the DCA graph (

Biodiversity and vegetation cover
Maximum species richness for all taxa increased from 9 in 0.0001 m² to 103 in 100 m² (Table 2). Across grain    (Tables 2  and 3). At the association level, the richness patterns were more diverse (Table 4, Figure 6): While species richness of vascular plants in all associations of orders A (rocky) and B (xeric) were equally low (around 20 in 10 m²), in order C (meso-xeric) they ranged from 27 in C.1 to 54 in C.5. Mean richness of non-vascular taxa in 10 m² varied strongly within the orders, ranging from 0.3 to 16.6 in order A and from 1.3 to 9.6 in order C. Herb layer cover was generally higher in the meso-xeric communities (order C) compared to the xeric and rocky ones (orders A and B) (Tables 3 and 4, Figure 6), while cover values of non-vascular taxa varied independently of the order and were largely consistent with the respective richness patterns ( Figure 6). While we found the highest mean species richness of vascular plants in the mountain meadows of C.5

Site conditions
According to all ecological indicator values except soil reaction, orders A and B were indistinguishable, but different from the meso-xeric order C, whose species composition stands for less light, lower temperature, lower continentality, more moisture and more nutrients (Table 3, Figure 7). By contrast, EIVs for soil reaction were high in order A, low in order B and intermediate in order C (Table 3, Figure 7). In general, measured topographic and soil variables showed only weak or no systematic difference between the orders, except skeleton content, which was clearly higher in the rocky grassland order A (Table 3).
At the association level, there were also strong differences for most of the measured variables (Table 4), with the most prominent ones shown in Figure 8: Within both the rocky and the meso-xeric order, the associations were segregated along an elevation gradient, with A.1 and C.3 occurring at the lowest and A.4 and C.5 at the highest elevations. Rock and stone cover generally was higher and more variable in the associations of the rocky and xeric orders (except A.3) compared to the meso-xeric associations. Soil depth in most communities was low (around 10-15 cm), but much higher in C.3 and C.5 (around 30 cm). Soil pH was high in most of the associations (6.9-7.4), but lower in C.4 (6.3) and much lower in C.2 (4.7). Nutrient concentrations (N and P) were generally low, with the noticeable exception of association A.2 with approximately two times higher values than all other associations.

Description of the associations and communities
In the following paragraphs, the ten accepted vegetation units at association level are described, and their syntaxonomic position is discussed. Four units could be unambiguously assigned to described associations, while the six remaining ones are labelled with informal community names. Figure 6. Comparison of species richness and cover of the two taxonomic-functional groups among the plots of the 10 communities at association rank. Letters represent homogeneous groups (at α = 0.05) according to Tukey's posthoc test following a significant ANOVA.

A.1 Armerio elongatae-Potentilletum arenariae
(relevés 1-4 in Suppl. material 2, Table 1, Figure 9a) Characterisation: Rocky grasslands on serpentine bedrock with low vegetation cover (48-80%). They are dominated by Festuca pallens and Potentilla incana with Carex humilis, Dorycnium germanicum and Thymus praecox as occasional subdominants. Among the species diagnostic or constant for this association, Asplenium cuneifolium, Erysimum sylvestre, Koeleria pyramidata var. pubiculmis, Notholaena marantae and Silene otites were recorded in our relevés. The cryptogam layer covered 1% to 10% of the surface and was dominated by Rhytidium rugosum and Weissia brachycarpa. The latter species, together with Fissidens dubius and Mannia fragrans, belongs to the diagnostic species of the association in our data set.
Ecology and distribution: Our relevés were recorded on very steep slopes with predominantly southern aspect and elevation between 785 and 830 m. The rugged terrain had a high cover of rocks and gravel on the surface and very shallow soils rich in skeleton. The high soil pH is typical for serpentinite (in this case it corresponds quite well with mean EIV-R of 7.2). N and P are rather low, but organic matter is high, even if there is a strikingly low litter cover (Table 4). Most of these species are adapted to the high magnesium soil content of the serpentine habitats. There is no management on these steep and dry southern slopes, although there is some moderate grazing by  Rupicapra. Nevertheless, there is no litter accumulation, and the low productivity seems in balance with the humification, as shown by the high organic matter content. In this cluster, we found the highest number and cover of red-list species, although the total species number was low. The special weathering of the serpentinite, which leads to very steep slopes, the dark soil surface caused by the high organic content, the low soil depth, the high soil magnesium content, and the position at a narrow section of the valley that increases wind velocity, interact to a very dry and specific habitat type. The history of this habitat might date back to the Neogene, as indicated by the occurrence of the palaeoendemic Sempervivum pittonii (Magnes et al. 2020), which has its closest relative in Bulgaria (Lippert 1995). The association is probably endemic to Styria, besides Gulsen occurring also near Kirchdorf und Traföß ("Tragöß") about 30 km downstream the Mur river where the serpentine layer reaches the surface again (Mucina and Kolbek 1993). Syntaxonomy: Our relevés are from the locus classicus of this association, which was described by Braun-Blanquet (1961) and included in the alliance Stipo-Poion xerophilae and in the order Festucetalia valesiacae. Mucina and Kolbek (1993) described a new alliance Avenulo adsurgentis-Festucion pallentis within the Stipo-Festucetalia pallentis, based on the Armerio-Potentilletum as holotype. This alliance was intended to include all dry grasslands on serpentine in Austria. However, the alliance name is invalidly published as the name-giving Avenula adsurgens is not present in the original diagnosis (Art. 3f ICPN, Theurillat et al. 2021).   (Table 4). The community is dominated by Carex humilis and Festuca rupicola with a constant occurrence of Allium lusitanicum, Fallopia convolvulus, Origanum vulgare, Stachys recta, Teucrium chamaedrys, and Verbascum chaixii subsp. austriacum. The endangered Stipa styriaca reached its highest cover in this association.

C.5 p-value (n = 4) (n = 3) (n = 2) (n = 5) (n = 15) (n = 8) (n = 5) (n = 7) (n = 5) (n = 13) Species richness
Ecology and distribution: The sampled stands were not managed. They have deep soils, the second highest organic matter content among all clusters and also relatively high litter cover, P and N ( Syntaxonomy: This unit perfectly matches the association Phleo-Pulsatilletum nigricantis described by Braun-Blanquet (1961) from various locations in Carinthia and one location in Styria. As in the previous unit, Braun-Blanquet included this association in the alliance Stipo-Poion xerophilae. Following a proposal by Franz (1988), Mucina and Kolbek (1993) merged the Phleo-Pulsatilletum nigricantis with the Potentillo puberulae-Festucetum sulcatae Br.-Bl. 1961 and transferred it to the alliance Festucion valesiacae. However, while the Potentillo puberulae-Festucetum sulcatae has indeed a strong affinity to the Festucion valesiacae, the Phleo-Pulsatilletum nigricantis clearly belongs to the Stipo-Festucetalia pallentis. We therefore suggest maintaining the Phleo-Pulsatilletum nigricantis as a separate association, which probably includes all rocky grasslands on relatively well-developed soils in the Styrian and Carinthian part of the Austrian central Alps. As in other regions of Central Europe, these moderate rocky grasslands have a more or less closed herb layer and are dominated by Carex humilis and Festuca rupicola (Janišová and Dúbravková 2010;Willner et al. 2013). The exact delimitation towards the Seselietum austriacae Br.-Bl. 1961, which comprises the more extreme, open rocky grasslands with Festuca pallens in the same region, requires further studies.

A.3 Medicago minima-Melica ciliata community
(relevés 8-9 in Suppl. material 2, Ecology and distribution: The shallow soils on the steep south-facing upper slopes show a high skeleton content, high pH and P, but low N and organic matter, although the litter cover is very high (Table 4). Both sites were not managed. One is overgrown by shrubs (Berberis vulgaris, Euonymus europaeus, Prunus spinosa), and the other one has a strongly developed litter layer covering 80% of the soil surface. The number of species is low in all categories. We recorded this community in the Upper Inn valley, in FL and LA.
Syntaxonomy: This unit does not match any described association from the Eastern Alps. The closest syntaxon from the literature is the Astragalo-Brometum stipetosum capillatae of Kielhauser (1954), which was raised to association rank by Mucina and Kolbek (1993) under the name Achnathero-Stipetum capillatae.

A.4 Koelerio pyramidatae-Teucrietum montani
(relevés 10-14 in Suppl. material 2, Table 1, Figure 9f) Characterisation: Rocky grasslands on extremely shallow soils on calcareous slate ( Ecology and distribution: Most of the surface (30-90%) is covered by rocks or gravel. Litter cover and organic content are low, like N and P content, with a high pH (Table 4). The plant layer is open, covering between 40 and 60%, and it is quite species poor, containing only 16 to 19 vascular plant species in 10 m 2 . The cryptogam layer is well-developed, covering between 40 and 60% of the soil surface, and rich in species (median species number is 17, maximum 23 including both bryophytes and lichens), being the highest values of all clusters. In the past (until several decades ago) most of these grasslands were grazed by sheep. The lichen Thamnolia vermicularis subsp. subuliformis, which has its preferred habitat on alpine wind-heathland on stony open soil (Wirth et al. 2013;Nimis et al. 2018), is a witness for transport of thallus fragments by sheep because this wormlike fruticose chalkwhite lichen never develops vegetative propagules such as isidia or soredia. Recently, all of these grasslands have been moderately grazed by cattle. We recorded the association in Obermauern (OM) in the Virgen valley (East Tyrol) on the top and southern and southeastern slopes of a hill called "Burg", at elevations of about 1,400 m.
Syntaxomy: This unit corresponds to the association Koelerio pyramidatae-Teucrietum montani, first proposed by Franz (1988) and validly described in Mucina and Kolbek (1993). While Franz (l.c.) originally placed the association in the Festucion valesiacae, the latter authors classified it in the alliance "Diantho lumnitzeri-Seslerion albicantis". The community has some similarity with the Teucrio-Caricetum humilis Br.-Bl. 1961, which is tradition-ally classified in the Stipo-Poion xerophilae and includes rocky grasslands in the Upper Inn valley (Braun-Blanquet 1961;Mucina and Kolbek 1993). However, according to the species list in Franz (1988) and our own relevés, Carex humilis seems to be absent in this association. The floristic relationship between Teucrio-Caricetum humilis and Koelerio-Teucrietum montani should be further addressed in future studies.

B.1 Sempervivum tectorum-Festuca valesiaca community
(relevés 15-29 in Suppl. material 2, Table 1, Figure 10a) Characterisation: This unit was documented from the Upper Inn valley, between 983-1,295 m on steep slopes. The rather low litter cover is connected with a high number and cover of cryptogams, indicated by the saxicole species Grimmia ovalis, G. laevigata and Hedwigia ciliata occurring on the rocks and Crossidium squamiferum and Tortula ruralis as bryophytes from open dry soil, interwoven by the common pleurocarpous dry grassland species Abietinella abietina and Rhytidium rugosum. Among grasses, Festuca valesiaca has a high constancy and cover. A fairly high number of annual species, typical of Sedo-Scleranthetea, form the largest bulk of diagnostic species.
Ecology and distribution: Shallow soils with rather low litter cover and organic content, and low N and P (Table 4). The community was recorded only in the Upper Inn valley (KA, FL and LA).
Syntaxonomy: This cluster is transitional between the orders Stipo-Festucetalia pallentis and Festucetalia valesiacae. In five relevés, the diagnostic species of the first order are more abundant, in seven relevés those of the second order, while in three relevés both species groups have more or less the same total cover (Figure 4). Diagnostic species of the order Festucetalia valesiacae that are present in this unit are Bothriochloa ischaemum, Festuca valesiaca, Medicago minima, Stipa capillata and Trifolium arvense. Kielhauser (1954) included similar grasslands of the same region in the Astragalo-Brometum. However, since the Astragalo-Brometum is a Brachypodietalia community (see below), this association name cannot be applied to unit B.1. Here, we provisionally include this unit in the alliance Festucion valesiacae. Without a comprehensive revision of all inner-alpine dry grasslands it is impossible to say if it might be assigned to one of the numerous associations described by Braun-Blanquet (1961) from South Tyrol and Switzerland, or if it must be described as a new association.

C.1 Astragalo onobrychidis-Brometum erecti
(relevés 30-37 in Suppl. material 2, Figure 10b) Characterisation: Grasslands in the Upper Inn valley dominated by Bromus erectus and/or Brachypodium rupestre, Carex humilis and Festuca rupicola. This unit is similar to the previous one, but it has a much lower cover of stones and deeper soils. Festuca valesiaca, Sempervivum tectorum and other diagnostic species of unit B.1 are rare or absent.
Ecology and distribution: Our plots have open and shallow soils, which are poor in skeleton, organic matter, N and P, with a rather high litter cover (Table 4). Braun-Blanquet (1961) considered the association as typical for the calcareous Bündner schist of the geological unit "Lower Engadine Window", from Ardez in Switzerland to Landeck in Austria.
Syntaxonomy: This cluster can be assigned to the Astragalo-Brometum, described from the Lower Engadine (Switzerland) by Braun-Blanquet (1949) and originally included in the Stipo-Poion xerophilae. Braun-Blanquet's alliance assignment was followed by Mucina and Kolbek (1993). However, the type relevé of the association (selected by Terzi et al. 2017) clearly belongs to the Brachypodietalia pinnati. Likewise, the Brachypodietalia species prevail in most relevés of this cluster, although the Stipo-Festucetalia pallentis species also have a high proportion (Figure 4). We therefore follow the interpretation of Willner et al. (2019) who included the Astragalo-Brometum in the Cirsio-Brachypodion. Relevé 49, which contains almost no Brachypodietalia species, is probably misclassified. Kielhauser (1954), who studied the dry grasslands in the Tyrolian Upper Inn Valley, distinguished two subassociations within the Astragalo-Brometum: the A.-B. stipetosum capillatae on steep, rocky slopes (see unit A.3), and the A.-B. festucetosum valesiacae on less rocky or even deep soils. However, the latter subassociation is a heterogeneous unit, including both xeric grasslands of the Festucetalia valesiacae (corresponding to our unit B.1) and meso-xeric (semidry) grasslands of the Brachypodietalia (corresponding to the Astragalo-Brometum s.str. and our unit C.1).
Ecology and distribution: This unit occurs on siliceous soils (significantly lowest pH of all clusters, Table 4) over quartzite bedrock north of Neumarkt in der Steiermark (NM). Litter cover and organic matter are high (the latter is the highest of all clusters, Table 4), a consequence of abandonment as the grassland management was readopted just the year previous to our sampling, and also the high P and N contents are indicative for accumulation processes. The bryophytes Bryum argenteum and Ceratodon purpureus indicate high nutrient levels.
Syntaxonomy: This community has some similarity with the associations Chamaecytiso hirsuti-Koelerietum and Genisto sagittalis-Festucetum rupicolae, both (invalidly) described from southern Styria -near to the Slovenian border -by Steinbuch (1995). In the revision of the semidry grasslands of Central and Eastern Europe by Willner et al. (2019), the Chamaecytiso hirsuti-Koelerietum was partly assigned to the Festuco rupicolae-Brometum Zielonkowski 1973, and partly to the Colchico-Festucetum rupicolae Lengyel et al. 2016 while the Genisto-Festucetum rupicolae remained unclassified, indicating that there are still some unresolved problems in the classification of semi-dry grasslands in Austria. There is also a slight similarity to the Brachypodio-Avenuletum adsurgentis that was (also invalidly) described from the Köszeg Mts. in western Hungary (Kovács 1994). Here, we prefer to use an informal community name for this cluster, leaving the clarification of its syntaxonomic status to a future revision.
Ecology and distribution: This cluster mainly includes the semi-dry grasslands on base rich, very deep and fine soils near Pöls (PÖ) that had been abandoned for years before grassland management was started again by clearing shrubs only a few years ago. Because of the occurrence of Stipa styriaca, the site was declared as a Natura 2000 protected area. With the reintroduction of grazing by sheep in 2020, there is a good chance that Stipa will be able to spread again from the rocky areas (see unit A.2) to the parts with deep soils. In this community, there is nearly no rock and gravel cover and also skeleton soil content, P and N are low (Table 4). The high litter cover does not lead to high organic matter, but suppresses the development of a bryophyte layer as well as the germination of most vascular plant species. The one relevé from Griffen (variant a) represents a grassland parcel within the bounding walls of the castle that was cleared from trees a few years ago, and the grassland management was started only after that. One relevé from Tyrol (KA) is included in this cluster as well.
Syntaxonomy: Three geographical variants can be distinguished in this cluster: The relevé from Griffen (variant a) is a strongly disturbed, ruderalized grassland of doubtful phytosociological position. The relevé from Kaunerberg (variant c) might represent a more mesic variant of the Astragalo-Brometum (see unit C.1). The remaining relevés (variant b) could be assigned to the Scabioso ochroleucae-Brachypodietum Klika 1933, which was described in Bohemia (Klika 1933). This association has not been reported from Austria before, but its presence in the Eastern Alps was predicted by the expert system developed by Willner et al. (2019). Against its Pannonian vicariant Polygalo majoris-Brachypodietum Wagner 1941, the Scabioso ochroleucae-Brachypodietum is only negatively differentiated.
Ecology and distribution: The soils are shallow and pH is remarkably low (with the exception of variant c) and both litter cover and organic soil matter are low (Table 4). P is low, while N has high values, especially in variant (b). The bryophyte layer with Calliergonella cuspidata, Plagiomnium undulatum and Thuidium philibertii shows a good water supply. The cluster includes relevés from three locations from each of the investigated main valley systems (NM, MA, FL).
Syntaxonomy: This is a rather heterogeneous cluster consisting of three geographical variants. Variant (a) includes a single relevé from Neumarkt in der Steiermark (NM), which is dominated by Arrhenatherum elatius and might belong to the alliance Arrhenatherion (class Molinio-Arrhenatheretea). Variant (b) contains two relevés from the Upper Inn Valley (FL), which have a high portion of acidophilous species (e.g. Calluna vulgaris, Danthonia decumbens, Viola canina). These relevés have some similarity with the Agrostio-Dianthetum deltoidis, which was described by Braun-Blanquet (1976) in the Lower Engadin and the Val Müstair (Switzerland). Variant (c) comprises two relevés from Virgen (MA) that share some species with the next unit.
Ecology and distribution: Grasslands on deep soils with low organic matter and low P and N (Table 4), mainly found in the Virgen valley. The subalpine plots from ZS and FL represent subalpine hay-meadows. The montane grassland in MA was managed similarly, mown only once and late in the year (August) without any fertilization. Although the subalpine grassland from ZS, an avalanche corridor, is not mown anymore, litter cover is low in all plots. Together with cluster 10b (part of C.4) we found here the highest vascular species number.
Syntaxonomy: This community is characterised by the presence of several subalpine and alpine species and therefore is transitional towards the class Elyno-Seslerietea. In this respect, it is similar to the Carlino-Caricetum sempervirentis Lutz et Paul 1947 of the northern Alps. According to the alliance concept of Willner et al. (2019), the Carlino-Caricetum sempervirentis belongs to the Mesobromion, while the grasslands in East Tyrol probably belong to the Cirsio-Brachypodion. However, the diagnostic species of the latter alliance are not very abundant in this cluster, and in some relevés they are even completely absent. Thus, the alliance assignment remains provisional. Another association that has similarity with this cluster is the Centaureo-Gentianetum cruciatae, described by Braun-Blanquet (1976) from the Upper and Lower Engadine (Switzerland). Some relevés from the highest elevations might already belong to the Campanulo scheuchzeri-Festucetum noricae Isda 1986(alliance Caricion ferruginei, see Grabherr et al. 1993, that was mentioned from other alpine hay meadows in the Virgen valley (Isda 1986).

Alliances and orders
While the assignment of the TWINSPAN clusters to phytosociological orders was relatively straightforward, the identification of the alliances remains doubtful for the most part.
The rocky grasslands of the Eastern Alps were included in three alliances by Mucina and Kolbek (1993): Avenulo adsurgentis-Festucion pallentis (comprising the Armerio elongatae-Potentilletum arenariae as the only association), Diantho-Seslerion (comprising the rocky grasslands on calcareous soils from East Tyrol and Salzburg eastwards) and Stipo-Poion xerophilae (comprising all dry grassland of the Inn valley). However, none of these names is acceptable because (1) the Avenulo adsurgentis-Festucion pallentis is not validly published (see description of unit A.1), (2) the Diantho-Seslerion is based on associations of the Bükk mountains in Hungary that are very different from the rocky grasslands of the Alps (see remark in the syntaxonomic scheme below), and (3) the Stipo-Poion xerophilae has been typified with a Festucetalia valesiacae community of the Vinschgau (Terzi et al. 2017). Thus, it seems that the rocky grasslands of the Eastern Alps are lacking a valid alliance name, unless they are included in a broadly conceived Seslerio-Festucion pallentis Klika 1931. However, it is impossible to decide from our dataset whether all rocky grasslands belong to the same alliance, or if they can be included in one of the alliances described from the Pannonian Basin. Therefore, we refrain from assigning our Stipo-Festucetalia pallentis communities to an alliance and leave this question to a broad-scale syntaxonomic revision of the order.
Grass steppes of the Festucetalia valesiacae are represented by a single community in our data set (Sempervivum tectorum-Festuca valesiaca community), which probably can be assigned to the Festucion valesiacae. It must be noted, however, that the current concept of the order Festucetalia valesiacae is much narrower than that of Braun-Blanquet (1961) who included all dry and semidry grasslands of the study area in this order. Whether the Stipo-Poion xerophilae can be maintained as an inner-alpine geographical vicariant of the Festucion valesiacae, remains to be further investigated.
Among the semi-dry grasslands of the Brachypodietalia pinnati, the communities of Styria (Agrostis capillaris-Avenula adsurgens community, Anthericum ramosum-Brachypodium pinnatum community) clearly belong to the Cirsio-Brachypodion, which includes the meadow steppes of eastern Central Europe (Willner et al. 2019). The most important differential species against the western Bromion erecti (= Mesobromion nom. cons. propos.) is Festuca rupicola, which was present with high cover in all relevés. The position of the Ranunculus bulbosus-Festuca rubra community and the Carduus defloratus-Brachypodium pinnatum community, both observed in the Virgen valley, is less clear. While they do not contain any diagnostic species of the Mesobromion (except for a single, probably misclassified relevé from the Inn valley where Festuca guestfalica was present), the Cirsio-Brachypodion species are also absent in half of the relevés. In fact, the Carduus defloratus-Brachypodium pinnatum community is transitional towards alpine grasslands of the Caricion ferrugineae (Elyno-Seslerietea), and so it even has a marginal position within the class. Finally, the Astragalo-Brometum of the Upper Inn valley, originally included in the heterogeneous Stipo-Poion xerophilae by Braun-Blanquet (1961), might belong to a hitherto undescribed meso-xeric alliance having its main distribution in the inner-alpine valleys of the Western Alps (and the dealpine gravel plains north and south of the Alps), differentiated against both Cirsio-Brachypodion and Mesobromion by the high constancy of Brachypodium rupestre and Potentilla pusilla (see Table 1), which are completely absent from typical stands of either alliance (Dengler et al., in prep.;see Dengler et al. 2020a).

Associations
Perhaps surprisingly, only four clusters could be unambiguously assigned to described associations. The main reason is that a comprehensive syntaxonomic revision of the dry and semi-dry grasslands of the inner Eastern Alps is still lacking. Instead of adopting names of associations that might or might not correspond to our units, we preferred to use informal community names, thus making the insufficient knowledge clearly visible. The relationship between these communities and described associations such as the Achnathero-Stipetum capillatae and Teucrio-Caricetum humilis (Mucina and Kolbek 1993) could only be clarified in a revision based on all available plot data from the Eastern Alps. For the Brachypodietalia pinnati communities, it will be particularly important to include data from Italy and Switzerland, which both were not covered by the revision of Willner et al. (2019). For the syntaxonomic interpretation of the Festucetalia valesiacae, a comparison with data from the most continental part of the Eastern Alps, the Vinschgau in South Tyrol (Italy), will be crucial.
Beside the lack of comprehensive studies another reason for the lacking match with associations from the literature might be the unstable management of most of the sites during the last thirty years. Even at sites that have recently become protected under the Natura 2000 regime (like the grasslands of FL and KA in the Upper Inn valley or PÖ in the Mur valley), the current management is very different from the historical, much more intensive use, and it will take quite some time for the vegetation to get in balance with the new type of management.
Our results do not support a closer relationship of the dry and semi-dry grasslands of the Austrian inner-alpine dry valleys with the Pannonian steppes (see Willner et al. 2013), a finding that is also supported by genetic analyses of plant and animal species known from both regions , see paragraph nature conservation below).

Biodiversity
For vascular plants, our findings that meso-xeric stands were substantially richer in species than either rocky or xeric grasslands across all grain sizes, is consistent with results from various other regional studies (Dengler et al. 2012Pedashenko et al. 2013) as well as a synthesis across the Palaearctic (Dengler et al. 2020b). Our scale-dependent richness values were similar to mean richness from nested-plot sampling in meso-xeric grasslands across the Palaearctic, but about 25% lower in the xeric grasslands and about 40% lower in the rocky grasslands (GrassPlot Diversity Explorer v.2.10; https://edgg. org/databases/GrasslandDiversityExplorer; Biurrun et al. 2021). By contrast, our values for the first two orders (20.0 species for rocky grasslands, 26.9 species for xeric grasslands in 10 m²) were rather similar to richness data found for such types in the inner-alpine valleys of Italy (Wiesner et al. 2015: 25 species in 10 m²) and Switzerland (Dengler et al. 2019: 26.3 species for rocky and 24.3 species for xeric grasslands in 10 m²). Actually, if we exclude the high-elevation associations (transitional to Elyno-Seslerietea) of the meso-xeric grasslands also there the richness values were below average and similar to those found in Aosta Valley and Valais (Wiesner et al. 2015;Dengler et al. 2019). While for the Austrian stands one might be tempted to explain the low plot-scale richness with the small size of the majority of remaining patches, this is not the case for the two other regions where there are still large and well-managed dry grasslands persisting. Therefore, we assume that it is rather due to an impoverished species pool in consequence of the glaciations and the isolated position compared to other dry grassland types, but this definitely requires further studies in a broader geographic context, for which the GrassPlot database provides excellent opportunities . By contrast, the strong increase of the plot-scale vascular plant species richness with elevation in the meso-xeric order (Table 4, Figure 6: from 27 to 54 species in 10 m²) is a well-known phenomenon for grasslands of the Alps. Data from the Swiss Biodiversity Monitoring, for example, also show that mean vascular plant species richness across all grassland types increases from 28 via 33 to 44 species in 10 m², in the colline, montane and subalpine belts, respectively (Koordinationsstelle Biodiversitäts-Monitoring Schweiz 2009). Specifically, the high species richness in subalpine hay meadows is a well-known phenomenon (Isda 1986;Matouch et al. 2000;Mayer and Grabner 2004). For bryophytes and lichens, the situation was reversed in comparison to mean values from nested-plot sampling throughout the Palaearctic (GrassPlot Diversity Explorer v.2.10; https://edgg.org/databases/GrasslandDiversityExplorer; Biurrun et al. 2021). Here, the Austrian stands of the rocky and xeric order were on average nearly twice as rich in species than the Palaearctic average, while in the meso-xeric order we had fewer bryophyte species normally with unclear pattern for lichens in the meso-xeric order. One possible explanation is the large species pool of terricolous cryptogams of open stands from the nearby alpine habitats in case of the first two orders.

Nature conservation
Based on the similar floristic composition of their vegetation, traditional vegetation ecology has anticipated close ties between the inner-alpine dry valleys and the Eastern European steppes (Braun-Blanquet 1961). Such scenarios assume that the biota of inner-alpine steppes are relatively young remnants of the cold-stage steppe belt. In other words, the inner-alpine steppe vegetation was often seen as a young and species poor derivative of the steppe vegetation of Eastern Europe. Genetic data from multiple steppe plants and insects challenged this scenario: The biota of the inner-alpine dry valleys were shown to consist of genetic lineages that have evolved in long-term isolation, and are not closely related to European steppe biota from Central and Eastern Europe . It was also shown that the genetic separation between the biota of inner-alpine and Eastern European steppes did not occur postglacial, i.e. at the beginning of the Holocene, but dates back as early as the mid-Pleistocene epoch . Accordingly, these extrazonal lineages must have survived climate driven range contractions throughout the Pleistocene climate fluctuations, such as during the LGM and the mid-Holocene forest optimum, in isolation, that is in refugia apart from the Pannonian and Eastern European steppes. The palaeoendemic Sempervivum pittonii from Gulsen can support these findings. Practically, the isolated nature of Alpine steppe outposts and their previously disregarded genetic uniqueness in respect to the Eastern European steppes raises their conservation value significantly. Adequate management and skilled nature protection activities are key to preserve the inner-alpine steppes and their rich genetic diversity for the future. In this respect, the importance of even small remnants of species-rich grasslands should be pointed out: they are often the last resources for propagules of autochthonous species for restoration projects (Kiehl et al. 2010;Bischoff et al. 2018).

Conclusions and outlook
Our study provides the first phytosociological overview of dry and semi-dry grasslands of the Austrian inner-alpine valleys using new plot data since the seminal work of Braun-Blanquet (1961). While a comprehensive syntaxonomic revision remains a task for future studies, the investigated plots are important references of the current situation. It is obvious that in regions with a still predominantly traditional type of farming, like in the Virgen valley, or in larger continuous areas with a proper management, like in the Nature Park Kaunergrat (Upper Inn valley), biodiversity of grassland-specialists is higher than on smaller grassland parcels within a matrix of high-input agriculture. Regarding the number and cover of red list species, the big serpentinite outcrop Gulsen in the Styrian Mur valley is the leaderwithout any management, but with very special habitat conditions. With the exceptions of the serpentinite grasslands and former hay meadows on avalanche corridors, all the studied grasslands are depending on a proper management, otherwise they would undergo mostly a natural succession to forests. Many of the grasslands are not only hotspots of biodiversity but also a cultural heritage, some of them maybe for millennia, and therefore have an extraordinarily high conservation value. The results of this study support the importance of grazing animals for achieving conservation targets in species-rich grasslands, as most of the detected biodiversity records are located in grazed habitats.

Data availability
The plot data presented in this paper were recorded on the occasion of the 11 th EDGG Field Workshop 2018 in Austria. They are included in the Suppl. material of this paper and are stored in and available from the GrassPlot database (https://edgg.org/databases/GrassPlot; Dengler et al. 2018).