Family Legacy: Depicting Diversity-Elevation Relationships of Tropical Tree Communities

Vitor de Andrade Kamimura (  vitorkami@msn.com ) Universidade Estadual Paulista (UNESP) https://orcid.org/0000-0002-3276-5812 Gabriel Mendes Marcusso Universidade Estadual Paulista (UNESP) https://orcid.org/0000-0002-7520-2876 Gabriel Pavan Sabino Universidade Estadual Paulista (UNESP) https://orcid.org/0000-0003-1284-8781 Marco Antonio Assis Universidade Estadual Paulista (UNESP) Carlos Alfredo Joly University of Campinas UNICAMP https://orcid.org/0000-0002-7945-2805 Priscila de Paula Loiola Universidade Estadual Paulista (UNESP) https://orcid.org/0000-0001-8760-2618


Introduction
Changes in species assemblages are associated with bounded abiotic conditions (Soberón 2007;Scheele et al. 2017), in which resource constraints should drive species' distribution (Huston 1994; Wang et al. 2020). Elevational gradients are among the most prominent abiotic gradients affecting species diversity and mirror latitudinal patterns (Rahbek 2005). Alongside elevation, other environmental factors change, such as temperature, atmospheric pressure, UV-B radiation, precipitation, wind speed and seasonality (Körner, 2007). Among these abiotic factors, temperature and humidity are expected to be the most important factors related to plant distributions (Gentry 1988 Elevational gradients are natural 'laboratories' for studying processes that cause variation in species distribution patterns (Lomolino 2001;Körner 2007) and thus to understand spatial variation in species composition, or Beta-diversity (Fitzpatrick et al. 2013). Beta-diversity is mainly caused by species turnover, and it should increase with environmental variations, revealing effects of ltering and dispersal limitation (Qian et al. 2009;Fontana et al. 2020). Plant Beta-diversity is expected to be hump-shaped along the elevational gradient (but see Moradi et al. 2020), however, its responses might be positive or negative, considering singular groups or shorter elevational ranges (Fontana et al. 2020). Beta-diversity was depicted into phylogenetic and species beta-diversities, to reveal the role of turnover at different phylogenetic levels. Species turnover should be higher with greater elevational variation, and a stronger effect should be observed to species, than to phylogenetic beta-diversity. elevation? (ii) How does phylogenetic diversity and structure change along the elevation gradient? Is this pattern different considering basal group in the analysis? (iii) How does alpha-diversity change along the elevational gradient? Is this pattern similar for the whole community and for the ve richest families in the community? We hypothesized that the beta-diversity would show high turnover for both phylogenetic and species beta-diversity, which should be highly correlated to elevational changes (Condit et al. 2002;Elliott and Davies 2019). We expected to nd a close covariation between species and phylogenetic diversity (Sandel 2018), and an increase of phylogenetic clustering at higher elevations, due to stronger environmental ltering (Körner 2007;Zhang et al. 2016;Qian et al. 2020). Finally, alpha-diversity should vary unimodally along the elevational gradient for the whole community (Rahbek 2005;Eisenlohr et al. 2013), emerging from patterns of the richest families (see Kamimura et al. 2017, Massante and Gerhold 2020).

Study area
The study was based on sampling plots of the tree community along the elevational range in a tropical montane system in the Serra do Mar State Park (SMSP, 23° 22' S and 45° 05'W), southeastern Brazil. The permanent plots are in three Nuclei of the SMSP (Cunha, Picinguaba and Santa Virgínia). The climate ranges from, Cfa to Cfb (Köppen's classi cation), annual average temperature goes from 20.8 ºC to 16.7°C and average annual rainfall ranges from above 2,200 mm to 1,780 mm, respectively from the lowlands to the highlands. Rainfall is well distributed throughout the year (Alvares et al. 2013).

Sampling design and data compilation
We complied and revised the forest plot dataset, with 17 plots of 10,000 m² (1 ha) each, used as local communities. All tree individuals with diameter at breast height (DBH) In the analysis, we only included individuals identi ed to the species level (c.a. 86% of the total abundance), excluding dead individuals, morphotypes (not revised by specialists) and conferatum (cf.) or a nis (aff.) taxa.

Data analysis
We assessed the beta diversity by means of phylogenetic and species diversity similarities decay. Phylogenetic beta diversity was the fraction of branch-length shared between communities, using phylosor in the picante library (Kembel et al. 2010). Species beta-diversity was calculated using the Sørensen index of similarity, with beta.pair in the betapart package (Baselga et al. 2018), partitioned into the turnover and nestedness-resultant components. We then assessed correlations among elevational divergence with phylogenetic and species beta diversities through Mantel Partial tests (Legendre et al. 2005), to account for spatial autocorrelation, using mantel.partial in the vegan package (Oksanen et al. 2019). Finally, we measured the goodness-of-t of models through multiples regression on distance matrices, based on permutation tests, using MRM in the ecodist package (Goslee and Urban 2007).
Phylogenetic diversity and structure were analyzed under two scenarios, including or not the tree ferns, the most distantly related clade (Cyatheaceae, in our case). We built two phylogenetic trees, considering the whole community and excluding Cyatheaceae species (ferns), from a consensus tree for seed plants, We then computed the phylogenetic diversity for each community. The phylogenetic diversity was expected to be correlated to species diversity (Sandel 2018), thus, we used a phylogenetic diversity index (PDI), which standardizes the phylogenetic diversity (Faith 1992)  We also computed alpha species diversity for (i) whole community (all species) and (ii) for the ve richest families in the study area (which comprised more than 50% of the total species richness). For each plot and for the selected families, we estimated species diversity in the matter of Hill numbers (Chao et al.

2014), reporting the Shannon's diversity index (q=1).
To detect the effects of elevation on phylogenetic diversity and structure, and on species diversity, we performed linear mixed-effects models (LMM) using the 'lme4' package (Bates et al. 2015). For each scenario analyzed (see above), we compared the results of the null model (diversity~1) and models with elevation or quadratic function elevation as xed effects. Forest types were used as a random factor to account for the different numbers of plots across forest types. The best models were chosen based on the Akaike information criterion (AIC). We tested the signi cance of each explanatory variable using the Anova function in the car package (Firth et al. 2009).
To account for spatial autocorrelation, we applied spatial autoregressive models (SAR; Bivand et al. 2013). Moran's I was calculated at 5-m interval distances between 0 and 50 m. Moran's I was signi cant at 5-10 m (P ≤ 0.001). Thus, we constructed the spatial weights matrix, in which the distance 10 km was de ned for the response variable (diversity) reached its maximum. We then added the spatial structure to the models and evaluated the goodness of t of the regression models by means of Nagelkerke R-Square (R²) (Nagelkerke 1991). Spatial autoregressive modeling was conducted by using the spautolm function implemented in the in library spatialreg (Bivand and Piras, 2015). All analyses were performed using R (R Core Team 2021), adopting α ≤ 0.05, and permutational tests were based on 999 randomized datasets obtained from Monte Carlo randomizations.

Beta-diversity along the elevation gradient
Both phylogenetic and species dissimilarities increased with increasing elevational divergence (Figure 1), and beta-diversity was mainly due to species turnover. Species turnover means were 0.86 for the phylogenetic and 0.91 for species approach. For both phylogenetic and species beta diversities, models presented similar coe cient of determination (R² = 0.53 for PD and R² = 0.45 for SD) and Mantel-r values ( Figure 1). Species dissimilarities were higher than phylogenetic dissimilarities following differences in the elevation of the plots, ranging from 0.5 to 0.75 to PD, compared to 0.25 to 0.5 of SD ( Figure 1).

Elevational patterns of phylogenetic diversity and structure
Correlation between species diversity and phylogenetic diversity was strongly affected by the choice of species scenarios in analyses (Figure 2a). For both species' scenarios, including or not the tree ferns, phylogenetic diversity monotonic decreased in with elevation ( Figure 2b). The inclusion of Cyatheaceae in the analyzes also affected the results of the phylogenetic structure (Figure 2c).
In the scenario 1, elevation affected the phylogenetic structure, driving a monotonic increase of NRI values along the gradient. The random phylogenetic structure was predominant among the plots (73% of the total), but in the Restinga forest, phylogenetic structure was clustered, while in the plots above 1.000 m a.s.l presented overdispersed phylogenetic structures. In the scenario 2, elevation slightly affected the phylogenetic structure, driving an inverted hump-shaped model of NRI variation along the gradient. The random phylogenetic structure was again predominant (94% of the total). A strong clustered phylogenetic structure was found in Restinga forest, while only one plot above 1.000 m a.s.l presented overdispersed structure (Figure 2c).

Changes in alpha species diversity along the elevation gradient
Overall, we found signi cant effects of the elevation on the species diversity patterns (Table 1). For the whole community, species diversity was unimodally distributed along the elevational gradient (Figure 3a, b). Regarding the richest families, three different variation patterns were found: (i) unimodal variation pattern (Myrtaceae); (ii) monotonic decrease in species diversity with elevation (Fabaceae); (iii) monotonic increase in species diversity with elevation (Lauraceae), besides insigni cant changes in species diversity for Melastomataceae and Rubiaceae (Table 1, Figure 3a, b).

Discussion
Evaluating diversity-environment relationships at different species scenarios is important to understand how tropical tree community patterns emerge from its different components. Phylogenetic and species turnover rates both increased with the elevational divergence between communities, and species dissimilarities were more pronounced than phylogenetic dissimilarities. Random phylogenetic structure occurred at low and intermediate elevations, becoming clustered at higher elevations. Also, assessing phylogenetic and species-diversity shed light on the processes shaping spatial patterns of biodiversity. We showed that diversity patterns differ for the whole community and for the richest families. Last but not least, since the effects of elevation on species distribution of rich families were different, we highlight the importance of assessing diversity-elevation for different species assemblages of tropical tree communities, especially when climate changes have profound impacts on tropical biodiversity of mountain systems.

Beta-diversity along elevational gradients in tropical forests
Beta diversity was mainly explained by turnover with a very low fraction of nestedness, both in phylogenetic and species terms. It is expected that species dissimilarity between communities increases with the geographical distance between them ( In line with our rst hypothesis, species and phylogenetic beta diversity were similar, whilst species turnover was higher than phylogenetic, across different elevations. On one hand, similar patterns of species and phylogenetic beta diversity found along environmental gradients can be associated with the conservatism of functional attributes (Du et al. 2021). On the other hand, species turnover was higher in relation to phylogenetic, indicating species diversi cation observed in species-level, rather than in genus or family levels. Temperature, resources and energy limitations along the elevational gradient, the harsh local conditions in stressful habitats, but not environmental threshold capacity or limited dispersed diasporas, have been described as determinants of species dissimilarities along elevational gradients (Qian 2009;Wang et al. 2020). Harsh local conditions may produce greater phylogenetic distance among regional abundant families in relation to less stressful habitats (Scarano 2002 3.2 Ecological processes driving local tree assemblages: species scenarios matter Using a phylogenetic approach, we found trends that have not explicitly been described (Zhang et  Three among ve richest families alpha diversities were related to elevation, however with different patterns, indicating an interaction between environmental drivers and distributions constraints conducting the diversity patterns (Colwell et al. 2016). Those differences also re ect the biogeographic history of these families of tropical wet and hot climates, which can help to explain the current patterns

Conclusions
By analyzing beta diversity under species diversity and phylogenetic diversity approaches, we demonstrated that elevation and local harsh conditions lead to higher species turnover, stronger at species than at family level. We show that the inclusion of basal plant groups had signi cant effects on phylogenetic analyzes of diversity, which mainly decreased along the elevation gradient, indicating an effect of environmental ltering in driving species assemblages. Phylogenetic and alpha diversities patterns differed, helping to understand the processes driving community assemblages. By splitting the analyzes of alpha diversity into different species scenarios, we showed that diversity-elevation relationships can be different for the richest families, whilst analysis of the whole community leads to the general unimodal pattern in the elevation gradient. Thus, we recommend merging ecological and evolutionary approaches, and yet splitting the analyzes into different species scenarios, to advance the comprehension of processes and mechanisms leading to diversity-environment relationships. Finally, in view of the differences of diversity variation patterns among rich tropical families and the delimitated spatial distribution of some phylogenetic lineages, that occur in their large majority at highest altitudes (i.e. Cunoniaceae and Winteraceae), we advocate that conservation planning also consider the diversity patterns of rich families to escape more biodiversity losses by the current and future climate scenario. Correlation between species and phylogenetic diversities (a) and variation across elevation in phylogenetic diversity (b) and phylogenetic structure (c). Phylogenetic diversity was computed by standardized measures of Faith´s diversity. Phylogenetic structure was assessed through standardized effect size of mean pairwise phylogenetic distances. Filled circles and triangles had signi cantly different phylogenetic diversity structure indexes in relation to the null model. cor -Pearson´s correlation; R² -Nagelkerke R-Square.