Exploring Plant Functional Diversity and Redundancy of Mediterranean High-Mountain Habitats in the Apennines

: We analyzed plant functional diversity (FD) and redundancy (FR) in Mediterranean high-mountain communities to explore plant functional patterns and assembly rules. We focused on three above-ground plant traits: plant height (H), a good surrogate of competition for light strategies, and speciﬁc leaf area (SLA) and leaf dry matter content (LDMC), useful indicators of resource exploitation functional schemes. We used the georeferenced vegetation plots and ﬁeld-measured plant functional traits of four widely spread vegetation types growing on screes, steep slopes, snowbeds and ridges, respectively. We calculated Rao’s FD and FR followed by analysis of standardized effect size, and compared FD and FR community values using ANOVA and the Tukey post hoc test. Assemblage rules varied across plant communities and traits. The High FR H registered on snowbeds and ridges is probably linked to climatic ﬁltering processes, while the high FD H and low FD SLA and FD LDMC on steep slopes could be related with underlying competition mechanisms. The absence of FD patterns in scree vegetation pinpoint random assembly processes which are typical of highly unstable or disturbed ecosystems. Improved knowledge about the deterministic/stochastic processes shaping species coexistence on high mountain ecosystems should help researchers to understand and predict vegetation vulnerability to environmental changes.


Introduction
During the last few decades, there has been considerable interest in ecological research to analyze the distribution and fitness of natural communities across environmental gradients [1] to understand community assembly rules, a key factor in assessing the effects of global change on biodiversity [2][3][4]. Previous studies explored the processes behind species assemblages by analyzing taxonomic diversity [5][6][7][8], with the assumption that all species are equally dissimilar and contribute in a comparable way to ecosystem functioning [9,10]. However, there is great functional variability among plant species and different taxa likely play different roles in ecosystem functioning [11,12].
Functional traits are morpho-physio-phenological characteristics that affect the fitness of species, i.e., their growth, reproduction, competition and survival, in a specific environmental context [13]. Given their importance in providing early signals of environmental change and depicting the relationships between species and ecosystem functioning, there has been an increasing number of plant community studies based on the analysis of functional traits [10,[14][15][16][17].
The coexistence of species in a community is determined by the interplay of deterministic processes (abiotic and biotic drivers) and stochastic processes (random drivers) [1]. Thus, functional diversity values (e.g., the variability of functional traits) of plant communities assembled by deterministic processes tend to depart from the diversity values that would be expected by chance [1,18]. In plant communities with high functional diversity values, species coexistence is ruled by biotic interactions such as competition. These interactions promote niche partitioning mechanisms ("limiting similarities") [19] and prevent the occurrence of species with similar traits, or, in other words, a species assemblage with low functional redundancy. In the short term, such plant communities marked by specialized and low-redundance taxa could be particularly vulnerable to environmental alterations, as the extinction of a single species and the attendant loss of its ecological role could alter the ecosystem's functioning [3,20]. Moreover, in the long term, high functional diversity values could support the ability of the altered plant community to adapt to environmental changes [11,21,22]. On the other hand, in a community with lower functional diversity than expected by chance, species coexistence may be ruled by harsh environmental factors that promote species convergence towards a few well-adapted and redundant functional traits in a process called "habitat filtering" [23]. Plant communities marked by species that share similar traits are quite resistant, because the loss of one species can be buffered by the presence of other functionally similar species, and thus the functioning of the ecosystem would not undergo alterations. However, in the long term, this low range of functional strategies would limit the altered plant community's ability to adapt to and survive further changes [9].
While species assemblage patterns such as niche overlap or partition and their relation to ecosystem functioning have been analyzed in grassland and forest ecosystems [15,24,25], there is a need for the analysis of ecosystems particularly affected by and sensitive to global change, such as the arctic tundra and high mountains [17,26]. There has been some research dealing with single aspects of plant diversity in high-mountain communities, such as taxonomic diversity [27,28] and functional diversity [29][30][31], but functional redundancy issues have yet to be explored sufficiently [32].
High-mountain ecosystems are hotspots of plant diversity [33][34][35] expressing a unique interplay of biogeographic, biotic and abiotic factors. Unfortunately, high mountains are threatened by direct and indirect human pressures that alter biodiversity and ecosystem functioning. In particular, Mediterranean summits can be considered prominent "hotspots" of climate change [35,36], where biodiversity erosion is highly concerning [37]. There is evidence that, on these summits, the rapid increase in temperatures and drought stress is altering the composition and function of plant communities [32,38,39]. As on other summits around the world, mountain ecosystems in the central Apennines are being altered by global climate change [35,38], and there is an urgent need for more in-depth research on functional strategies that cope with the ongoing changes. Understanding functional assemblage mechanisms as well as functional redundancy issues at the species and community level is crucial for modelling and forecasting how biodiversity will change and how the functioning of ecosystems will alter in response to these global changes [40].
Given these considerations, the present work sought to analyze species assembly and functional redundancy across high-mountain Mediterranean plant communities. To explore the underlying mechanisms of plant species' local coexistence, we analyzed functional diversity and redundancy based on a set of traits commonly used for depicting processes of competition (plant height: H) and resource exploitation (specific leaf area: SLA, leaf dry matter content: LDMC) that are involved in specific ecosystem services such as soil fertility and climate mitigation [41]. We specifically analyzed the ecosystem mosaic in the calcareous central Apennines, formed by four main vegetation types: the Leucopoa dimorpha community growing on screes, the Sesleria juncifolia community on steep slopes, the Plantago atrata community on snowbeds and the Silene acaulis community on ridges. In particular, we focused on the following questions: (1) How do functional diversity and

Study Area
We analyzed four plant community types representative of the high elevation zone of the Apennines limestone chain (Appendix A) and referable to the specific land units of screes, steep slopes, snowbeds and ridges [42]. Specifically, we considered: (a) the Sesleria juncifolia community, which consists of calciphilous stepped and garland grasslands, common on rendzina soils, sampled on steep slopes; (b) the Leucopoa dimorpha community that grows on limestone screes characterized by discontinuous vegetation growing on steep slopes with dry and debris-rich soils; (c) the Silene acaulis community represented by discontinuous vegetation in wind-scoured ridges growing on shallow soils, rich in debris; and (d) the Plantago atrata community, which consists of Oro-Apennine continuous grasslands growing on small snowbeds where there is a long period with long snow cover, a short vegetative period and humid soils [31,[42][43][44][45][46]. Table 1 reports information on taxonomy (species richness and number of species), environmental conditions (elevation and slope) and ecological conditions, quantified with Ellenberg indicator values (ordinal classification of plants according to the position of their realized ecological niche along an environmental gradient) [47].
All the selected plant communities are habitats of conservation concern in Europe (EUcode 6170: Alpine and subalpine calcareous grasslands, EUcode 8120: Calcareous and calcshist screes of the montane to alpine levels) [48]) and are particularly rich in endemics, Southern European orophytes and Mediterranean montane species [42,49,50].
The climate in the central Apennines shows a typical Mediterranean pattern with low summer precipitation, negligible summer aridity and a wetter period in autumn. Summer precipitation increases towards higher elevations, which have a humid/hyperhumid water regime [51][52][53]. According to the Köppen-Geiger classification, the climate ranges from temperate-oceanic (Cfb) in the lower sector to subarctic (Dfc) at the summits [54]. Table 1. Brief description of the analyzed plant communities (A: Leucopoa dimorpha community, B: Sesleria juncifolia community, C: Plantago atrata community, D: Silene acaulis community) occurring in the different land units (according to Stanisci et al. [42]) in terms of taxonomic diversity (species diversity reported as 1-Simpson), community mean for Ellenberg indicator values (CM EIVs ) for Italian flora [55] calculated with presence/absence species data (L, light; T, temperature; U, soil moisture; R, soil reaction; N, soil nutrients) and local environmental features registered at plot level. Mean plot values along with standard deviation (in brackets) are reported.

Data Collection
We extracted vegetation data from the VIOLA database (high mountain VegetatIOn of centraL Apennines [45,56]; European Vegetation Archive code-EU-IT-019 [57]). Plots carried out during the last few decades describing the four target plant communities were extracted and selected according to: (a) the presence of the pool of the diagnostic species, as reported in the Italian vegetation prodrome and EUcode manual for habitat identification [58,59], (b) maximized inter-plot spatial distance (plots separated by at least 200 m) and (c) comparable vegetation cover among plots. In total, 164 plots were selected, with an average size of 45 m 2 , distributed as follows: 23 of the Leucopoa dimorpha community, growing on calcareous screes; 55 of the Sesleria juncifolia community, occurring on steep slopes; 49 of the Plantago atrata community, growing in snowbeds, and 37 of the Silene acaulis community, growing on mountain ridges. Taxa nomenclature follows that of Pignatti [60].
For functional analyses, given that it was not feasible to measure the trait values for all species in each plot [61], we selected the most abundant species limited to those taxa whose relative cumulative cover was 70% of vegetation cover on at least one of the considered plots [61]. Thus, for each plot, we ordered the species by decreasing relative cover values and, beginning from the most abundant one, we summed them up to 70% of the total plot vegetation cover. Following this procedure, we identified a set of 62 species for which we measured vegetative plant height (H, cm), specific leaf area (SLA, mm 2 /mg) and leaf dry matter content (LDMC, mg/g), which are good surrogates of the ecological functions of competitive ability and resource exploitation [62]. Species trait values were partially retrieved from recent measurements performed in the analyzed area (2016-2017) [31] and by dedicated field measurements carried out following the standardized protocol (2019) [62], in line with Zanzottera et al. [63]. Trait research based on data specifically measured in the analyzed area ensures more accurate information about plant strategies than that obtained from databases [64].

Indicator Species Analysis and Species-Level Redundancy
First, we explored the distribution pattern of the 62 most abundant species and their measured traits (H, SLA and LDMC) across the four analyzed plant communities by implementing indicator species analysis (ISA) [65] followed by species redundancy analysis [66]. The "species x plot" matrix was analyzed using plant community types as a grouping factor, with 999 permutations [65]. We identified the species discriminating each plant community (hereafter "community indicator species"), as those with ISA relative abundance over 0.6 and relative frequency higher than 0.25 [67].
Next, for each of the community indicator species identified, we measured their individual trait redundancy value (r H , r SLA , r LDMC ) with respect to the other species of the plant community as follows [66]: where r i is the redundancy value of species i with respect to the other species in the community, and δ is the functional dissimilarity for a given trait between species i and j. Functional dissimilarity (δ ij ), calculated with Gower distance [68], is = 0 when two species assume exactly the same trait values and 1 when two species have exactly the opposite trait values. Species redundancy values are expressed in a range from 0 to 1. The ISA was performed in the R environment (R Foundation for Statistical Computing, Vienna, Austria, http://www.R-project.org, accessed on 1 April 2021) using the mutipatt function in the indicspecies package, while the species functional redundancy was calculated with R function uniqueness, provided by Ricotta et al. [66].

Community-Level Functional Diversity and Redundancy
The next step was to calculate the community-level functional diversity (FD) and redundancy (FR) using Rao's Quadratic Entropy (Q) [69]. We selected Rao's Q because it is able to summarize the abundance of the species and the functional distance of their traits and offers a sound framework for the simultaneous analysis of a variety of facets, such as taxonomic and functional parameters [70]. Rao's Quadratic Entropy expresses the expected dissimilarity between two individuals of a given assemblage selected at random with replacement: where S is the number of species, δ ij is the dissimilarity between the i-th and j-th species, and p i and p j are the proportion of abundance of the i-th and j-th species in the vegetation plot. When Q is implemented for measuring FD, the dissimilarity (δ ij ) between the i-th and j-th species may vary from 0 (two species have the same trait values) to 1 (two species have completely different trait values).
Rao's Q was also used for calculating the functional redundancy (FR), as follows [66]: where FD is functional diversity and TD is the Simpson index of dominance (1-Simpson) calculated with Rao's Q. TD represents the maximum value that FD may achieve, since the distance between each pair of species is fixed as δ ij = 1 for all i = j. Thus, FD may vary from 0 (all species present the same trait values) to TD (all species are completely functionally different) [63]. FR can assume values ranging from 1 (a highly redundant community in which all species are functionally identical) to 0 (no redundant communities with all the species functionally dissimilar). It should be noted that since FR is a fraction of FD and TD, it is independent of species richness [66]. For each community (the Leucopoa dimorpha, Sesleria juncifolia, Plantago atrata and Silene acaulis communities), we calculated FR and FD for each trait (FR H , FR SLA , FR LDMC , FD H , FD SLA , FD LDMC ), using the Gower measure (δ ij ) because it standardizes the distance values (from 0 to 1) [68], allowing a comparison between TD and FD [70], which is necessary to calculate FR. Continuous trait values were log-transformed before calculation of the indices [61]. We calculated functional redundancy (FR) and functional diversity (FD) using the R function uniqueness [66].
To detect community assembly processes, we generated the distribution of expected random values of FD by shuffling the trait values across species 999 times and by calculating the standardized effect size [18] of FD for each trait (SES: observed FD values-mean 999 expected FD values/standard deviation of 999 expected FD values). Then, we assessed whether the distribution of SES values for each FD trait per plant community was significantly different from zero using a two-tailed t-test. Distribution of SES values significantly lower than 0 denotes "functional convergence", while that of values significantly higher than 0 indicates "functional divergence".
Variations in the standardized effect size of functional diversity (SES-FD) and in functional redundancy (FR) for each trait (i.e., H, SLA, LDMC) across the considered plant communities were quantified with distinct one-way analysis of variance (ANOVA) type I models, followed by Tukey post hoc tests [71]. For each significant model, the normality, homogeneity and independence of residuals were checked by the visual inspection of the residuals [72].
The aov, TukeyHSD and t.test functions in the stats package served to perform the ANOVAs, the Tukey post hoc test and the t-test, respectively. The mean values of species-level functional redundancy for plant height (r H ) were lower for the Sesleria juncifolia community (mean = 0.74) and the Leucopoa dimorpha community (mean = 0.75), due to the co-occurrence of different strategies, namely those of tall graminoids (Leucopoa dimorpha) and short forbs (Doronicum columnae). On the other hand, the mean value of species-level height redundancy (r H ) was higher in the Silene acaulis community (mean = 0.81; Pedicularis elegans and Galium magellense) and the Plantago atrata community (mean = 0.84, Poa alpina subsp. alpina, Ranunculus pollinensis and Trifolium thalii) ( Table 2). Table 2. List of community indicator species for each plant community derived from ISA analysis. For each taxon, we report the indicator value (IndVal) and the species-level redundancy value for each trait (r H : plant height; r SLA : specific leaf area; r LDMC : leaf dry matter content). Statistical significance of indicator values is also reported (n.s. p > 0.05; * p < 0.05; ** p < 0.01; *** p < 0.001). and high in the Sesleria juncifolia community (mean = 0.90; Sesleria juncifolia and Carex humilis). Intermediate values of r SLA were observed in the Leucopoa dimorpha community (mean = 0.82; with Achillea barrelieri subsp. barrelieri and Isatis apennina) and in the Silene acaulis community (mean = 0.86; Anthyllis vulneraria subsp. pulchella and Arenaria grandiflora subsp. grandiflora) ( Table 2).
The mean value of species-level redundancy for LDMC (r LDMC ) assumed lower values in the Leucopoa dimorpha community (mean = 0.65 with Leucopoa dimorpha and Isatis apennina) and higher values in the Sesleria juncifolia community (mean = 0.84; Carex humilis and Globularia meridionalis). Intermediate r LDMC values were observed in the species of the Plantago atrata community (mean = 0.73; Ranunculus pollinensis and Trifolium thalii) and of the Silene acaulis community (mean = 0.80; Androsace villosa subsp. villosa and Arenaria grandiflora subsp. grandiflora) ( Table 2).

Community-Level Functional Diversity
The

Community-Level Functional Diversity
The  Boxplot letters indicate statistically significant differences among communities after Tukey's post hoc test (p < 0.05).
Significant departures of SES-FDH from random expectation were observed for the Sesleria juncifolia community (functional divergence; Table 3) as well as for the Plantago atrata and Silene acaulis communities (functional convergence; Table 3). Conversely, the Leucopoa dimorpha community plant height assemblage did not differ from a random pattern (Table 3). Concerning leaf traits, the Sesleria juncifolia and Silene acaulis Significant departures of SES-FD H from random expectation were observed for the Sesleria juncifolia community (functional divergence; Table 3) as well as for the Plantago atrata and Silene acaulis communities (functional convergence; Table 3). Conversely, the Leucopoa dimorpha community plant height assemblage did not differ from a random pattern (Table 3). Concerning leaf traits, the Sesleria juncifolia and Silene acaulis communities showed a convergent pattern (negative departure from random expectation) for both SLA and LDMC (Table 3). Conversely, the Plantago atrata community presented a divergent pattern for SLA and a convergent one for LDMC (Table 3), while the Leucopoa dimorpha community distribution pattern of SLA and LDMC did not differ from a random expectation (Table 3). Table 3. Results of the two-tailed t-test assessing the standardized effect size of functional diversity per trait (SES-FD H : plant height; SES-FD SLA : specific leaf area; SES-FD LDMC : leaf dry matter content) and plant community (A: Leucopoa dimorpha community, B: Sesleria juncifolia community, C: Plantago atrata community, D: Silene acaulis community). Statistical significance of t-test is also reported (n.s. p > 0.05; * p < 0.05; ** p < 0.01; *** p < 0.001).

Community-Level Functional Redundancy
Functional redundancy (FR) significantly varied across the analyzed plant communities and presented complementary behavior with respect to FD (Figure 2; Appendix B).
We observed higher redundancy values for vegetative plant height (FRH) in the Plantago atrata and Silene acaulis communities (Figure 2a). Concerning leaf traits, the Sesleria juncifolia community showed a higher redundancy value for both SLA (FRSLA; Figure 2b) and LDMC (FRLDMC; Figure 2c). On the other hand, we registered lower redundancy values for SLA in the Plantago atrata community (FRSLA; Figure 2b) and for LDMC in the Leucopoa dimorpha community (FRSLA; Figure 2b).

Discussion
The distribution pattern of the 62 most abundant species varied across the different vegetation types. The relatively high number of indicator species for all the plant communities (higher number of species in the Silene acaulis community and lower number in Sesleria juncifolia communities) is most likely related to the high abiotic differences among the local environmental conditions of the plant communities forming the Apennine high-elevation vegetation mosaic [43]. Such high floristic variability between plant communities on high-mountain landscapes depicts the biodiversity response to topography, which plays a consistent role in modulating local climatic and soil conditions [73]. In the central Apennines, the interplay of climatic and edaphic factors across the altitudinal gradient has given rise to a complex mosaic of different vegetation types with We observed higher redundancy values for vegetative plant height (FR H ) in the Plantago atrata and Silene acaulis communities (Figure 2a). Concerning leaf traits, the Sesleria juncifolia community showed a higher redundancy value for both SLA (FR SLA ; Figure 2b) and LDMC (FR LDMC ; Figure 2c). On the other hand, we registered lower redundancy values for SLA in the Plantago atrata community (FR SLA ; Figure 2b) and for LDMC in the Leucopoa dimorpha community (FR SLA ; Figure 2b).

Discussion
The distribution pattern of the 62 most abundant species varied across the different vegetation types. The relatively high number of indicator species for all the plant communities (higher number of species in the Silene acaulis community and lower number in Sesleria juncifolia communities) is most likely related to the high abiotic differences among the local environmental conditions of the plant communities forming the Apennine highelevation vegetation mosaic [43]. Such high floristic variability between plant communities on high-mountain landscapes depicts the biodiversity response to topography, which plays a consistent role in modulating local climatic and soil conditions [73]. In the central Apennines, the interplay of climatic and edaphic factors across the altitudinal gradient has given rise to a complex mosaic of different vegetation types with distinct physiognomies and species composition [42] (see also Table 2). Accordingly, species coexistence patterns depicted by functional diversity (SES-FD H , SES-FD SLA and SES-FD LDMC ) and functional redundancy (FR H , FR SLA and FR LDMC ) varied across the analyzed vegetation mosaic. Functional redundancy varied but with a mirrored behavior with respect to functional diversity, confirming the "complementarity" of the information supplied by the two indices [74].

Vegetative Plant Height
The plant communities growing in particularly cold environments, such as the Plantago atrata vegetation of snowbeds and the Silene acaulis vegetation of ridges, presented significantly low functional diversity and high redundancy for plant height (low SES-FD H and high FR H ). We observed a convergence in plant height, which could be related to the stringent climatic filtering mechanisms of upper elevations [23,30,[75][76][77], which positively select short species able to benefit from a soil temperature that is higher than the air temperature [73,78]. Indeed, the climate of the Apennine summits and ridges is cold, with marked winter frosts [52] and long snow cover periods on snowbeds, all conditions that select functional strategies adapted to cold stress [29,31,76,77,79]-for example, those of specialist cryophilous species [42,78]. Here, the vegetation on snowbeds and ridges is dominated by short forbs (Trifolium semipurpureum: 3.1 cm; Ranunculus pollinensis: 2.7 cm; Anthyllis pulchella: 3.5 cm; Galium magellense: 2.46 cm; Pedicularis elegans: 4 cm) and short graminoids (Kobresia myosuroides: 8.5 cm; Festuca italica: 6.89 cm) The observed high redundancy for plant height (high FR H ) may enable plant communities to deal with environmental changes successfully in the short term, as the loss of one species could be buffered by the presence of other, functionally similar ones [9]. However, low plant height variability and high trait redundancy denote a reduced set of plant strategies, which indicates weak potential for reacting successfully to changing environmental conditions in the long term [11].
Conversely, mountain slopes with the Sesleria juncifolia community disclosed high values of plant height functional diversity (SES-FD H ) and low values of functional redundancy (FR H ). Given that vegetative plant height captures the functional dimension of competition for light [62], the high SES-FD H is probably related to niche partitioning processes (limiting similarities) [19] that regulate species coexistence in the high mountain stands, where the growing season is longer. Indicator species in the Sesleria juncifolia community showed great variability in plant sizes, ranging from tall graminoids (Sesleria juncifolia: 16.8 cm; Bromopsis erecta: 32.7 cm) to short forbs (Globularia meridionalis: 2 cm; Anthyllys montana: 5.46 cm). Here, the higher diversity may suggest that the vegetation has a good ability to adapt to climate change in the long term, since the presence of functionally dissimilar species should indicate the potential of the vegetation to adapt to new environmental conditions [9,11]. In addition, the good cover of tall species should contribute to climate regulation processes and local soil fertility services (Appendix C).
The random behavior of SES-FD in the Leucopoa dimorpha community growing in screes suggests an underlying stochastic assemblage mechanism. Indeed, screes are characterized by sparse vegetation, with a small number of glareicole species coexisting with species coming from the neighboring communities, which randomly migrate and temporally survive on such extremely unstable substrates [80]. As such, this may preclude the establishment of a predictable plant community functionally modelled by such non-random assembly rules as habitat filtering and biotic interaction [1].

Leaf Traits
For the Plantago atrata snowbed community, we recorded high SES-FD SLA values and low FR SLA , which, as observed in other mountain ecosystems [81], could be interpreted as a community response to locally milder soil conditions [64]. Indeed, snowbeds are morphological units distributed at different altitudes and characterized by relatively rich and humid soils (CM EIV for N; Table 1) [42], which may enhance local resource competition mechanisms and character shift strategies (Appendix C). Specifically, the high leaf diversity in snowbed vegetation is attributable to the co-occurrence of species having a broad spectrum of leaf functional performance ranging from fast-growing species (Taraxacum glaciale, SLA mean: 29.6 mm 2 mg −1 ) to slow-growing species (Plantago atrata, SLA mean: 10 mm 2 mg −1 ). Higher specific redundancy (functional similarity with coexistent species) was found for species having intermediate SLA values, such as Gnaphalium diminutum (SLA mean: 16.2 mm 2 mg −1 ) and Poa alpina (SLA mean: 18.4 mm 2 mg −1 ). Global warming and the consequent reduction in snow cover are expected to negatively affect snowbed ecosystems, with results such as alterations in biogeochemical cycles, reduced forage for chamois kids and the filtering of water that feeds springs of potable water [82,83]. Snowbed ecosystem changes may also cause short-term alterations in leaf functions, but the presence of functionally dissimilar plants might support good mid/long-term potential to cope with climatic and land use alterations.
In the Silene acaulis community growing on ridges, we noted the functional convergence of leaf strategies (low SES-FD SLA; LDMC and high FR SLA; LDMC ), which, like observations for plant height, could be related to a cold filtering effect. The higher redundancy of slow-growing species suggests the ability to hold the position in an extreme environment and to keep ecosystem functions after species loss (high resistance) [82]. However, the SLA values on criophilous indicator species are very variable and range from relatively high values for Galium magellense (SLA mean: 23.2 mm 2 mg −1 ) and Armeria majellensis (SLA mean: 19.42 mm 2 mg −1 ) to relatively low values for Anthyllis pulchella (SLA mean: 13.44 mm 2 mg −1 ), Silene acaulis (SLA mean: 12.68 mm 2 mg −1 ) and Festuca italica (SLA mean: 11.4 mm 2 mg −1 ). This high variability in species leaf performance is attributable to the cooccurrence of different growth rate strategies that range from acquisitive resource use (high SLA values), which optimizes rapid carbon gain during the short growing season [31,77], to conservative resource use (low SLA values), in which the adoption of a slow-growth strategy assures the ability to survive environmental hazards [73,79].
The convergent pattern registered for leaf traits (low SES-FD SLA; LDMC and high FR SLA; LDMC ) in the Sesleria juncifolia community growing on calcareous slopes could be related to the interplay of two filtering environmental factors, winter frost stress and soil drought [31], which are particularly severe in Mediterranean limestone massifs [30,52,84]. In particular, the slow growth of perennial species (grasses and sedges) represents a good adaptation for both freezing and aridity [29,85] (see also CM EIV for N in Table 1). This convergence towards similar leaf functional strategies well-adapted to climatic stress is seen in very similar indicator species (Aster alpinus mean SLA: 12.12 mm 2 mg −1 , and Sesleria juncifolia mean SLA: 9.50 mm 2 mg −1 ). The high functional redundancy of the Sesleria juncifolia community should support good immediate ecosystem resistance to environmental alterations [86] but, at the same time, the limited range of leaf strategies makes these grasslands poorly adapted to further long-term environmental alterations.
The leaf trait diversity of the Leucopoa dimorpha community of screes (SES-FD SLA; SES-FD LDMC ) did not differ from values generated by chance (Table 3), indicating the occurrence of casual processes of local colonization and extinction, and the absence of a mature, well-structured plant community. It should be noted that on high-mountain screes in the Central Apennines, soils are highly erodible due to the steepness of slopes, the incoherent substrate and the trampling pressure exerted by the endemic chamois [83]. The interplay of such dynamic abiotic and abiotic factors may enhance the stochastic colonization of plant species coming from neighboring habitats, as well as their local extinction.
In short, as observed in the temperate region [64], plant trait diversity in the central Apennine limestones seems to be related to local environmental variability.

Conclusions
The present research on high-mountain vegetation communities explored vegetation assembly rules from a functional perspective, as well as the degree of functional redundancy, and contributes to knowledge about the relations between plant species distribution patterns and the physical environment (snowbeds, ridges, steep slopes and screes) of Mediterranean landscapes. The interpretation of our results in light of previous research suggests the primary role of environmental features such as climate and soil moisture in community assembly, as we observed distinct selective effects (abiotic filters and biotic interaction) on all the measured above-ground plant traits and each considered plant community. Our findings gave further evidence of the essential contribution of trait-based analysis for understanding how ecosystems respond to environmental change. The varia-tion in functional diversity was coupled with a converse variation in functional redundancy, revealing that both indices capture complementary information-namely, the degree of species unicity vs. similarity.
In addition, the analysis of species co-occurrence and the assessment of species-level redundancy sheds light on the functional role of indicator species, which should be of particular concern for endemic taxa. The species-level analysis made it possible to identify those functionally similar taxa that better express the species coexistence at community level. Even so, further studies are needed to explore the effects of climate variation on these functional strategies both at species and community level.
The measured above-ground plant traits captured different aspects of the ecological niches of species [16], but further investigation using other sets of traits, such as belowground and reproductive traits, and incorporating intraspecific trait variability is needed to achieve a deeper understanding of the underlying mechanisms of species coexistence [77].   Figure A1. Location of the study area and the mountain massifs of the Central Apennines (Abruzzo and Molise Regions in white).

Appendix C. Functional Composition (Community Weighted Mean: CWMs) of the Analyzed Plant Communities for Plant Height (H), Specific Leaf Area (SLA) and Leaf Dry Matter Content (LDMC)
To better describe the functional characteristics of the compared communities, we used the community-weighted mean values for each single trait (CWM H ; CWM SLA ; CWM LDMC ), which summarize the mean values of traits at plot level [87]. This was computed by multiplying the "species x plot" matrix with the "species x traits" matrix as follows: where CWM is the community-weighted mean, x i is the trait value of a given species and p i is the relative frequency of the i-th species, based on cover data. We compared plant community CWM values using ANOVA followed by Tukey's post hoc test. In order to reduce inflated type I error rates [88], we integrated ANOVA analysis with the "Max Test" procedure [84]. We ran the ANOVAs integrated with the Max Test with the test.cwm function (method "aov", test "max") in the weimea package [89].
As reported in Figure A2, Plantago atrata is characterized by higher CWM SLA values ( Figure A2b) and lower CWM LDMC , which reflects its fast-growing strategies (higher cover of perennial forbs with competitive strategies), supported here by the fertile and humid soils of the snowbeds. The Sesleria juncifolia community presented lower values of SLA and higher values of LDMC, indicating the dominance of slow-growing strategies (higher presence of tall perennial grass species with competitive stress-tolerant strategies) able to grow and keep its position on high-mountain steep slopes. As previously reported [31], the Silene acaulis ridge community is characterized by low plants and high leaf dry matter content, denoting stress-tolerant mechanisms that increase their acquisitive strategies during the short growing season of the summer to maximize resource exploitation. Finally, the traits of the Leucopoa dimorpha community, growing on calcareous calcshit screes, depict random processes due to the dynamic substrate (see also the Discussion).
(higher presence of tall perennial grass species with competitive stress-tolerant strategies) able to grow and keep its position on high-mountain steep slopes. As previously reported [31], the Silene acaulis ridge community is characterized by low plants and high leaf dry matter content, denoting stress-tolerant mechanisms that increase their acquisitive strategies during the short growing season of the summer to maximize resource exploitation. Finally, the traits of the Leucopoa dimorpha community, growing on calcareous calcshit screes, depict random processes due to the dynamic substrate (see also the Discussion). Figure A2. Boxplots comparing the functional characteristics of the analyzed plant communities (A: Leucopoa dimorpha scree community, B: Sesleria juncifolia steep slope community, C: Plantago atrata snowbed community, D: Silene acaulis ridge community) in terms of community-weighted mean (CWM) for the functional strategies related to plant height (a), SLA (b) and LDMC (c). Boxplots with different letters indicate statistically significant differences among them after Tukey's post hoc test (p < 0.05). Statistical significance of ANOVA analysis is also reported (n.s. p > 0.05; * p < 0.05; ** p < 0.01). Figure A2. Boxplots comparing the functional characteristics of the analyzed plant communities (A: Leucopoa dimorpha scree community, B: Sesleria juncifolia steep slope community, C: Plantago atrata snowbed community, D: Silene acaulis ridge community) in terms of community-weighted mean (CWM) for the functional strategies related to plant height (a), SLA (b) and LDMC (c). Boxplots with different letters indicate statistically significant differences among them after Tukey's post hoc test (p < 0.05). Statistical significance of ANOVA analysis is also reported (n.s. p > 0.05; * p < 0.05; ** p < 0.01).