The impacts of climate change on the world’s ecosystems have already been documented on every continent, ocean and in most taxonomic groups (Scheffers et al., 2016). Greenhouse gases atmospheric concentrations are reaching levels never seen in recent history nor estimated over the past 20 million years (Beerling and Royer, 2011). Coupled with the intensification of habitat loss and aggressive land-use change, climate change represents one of the main threats to biodiversity and associated ecosystem services, particularly to tropical regions where most of the biodiversity is concentrated (Asner et al., 2010; Zwiener et al., 2017).

One of the most species-rich and yet highly degraded tropical domains is the Atlantic Forest (Mittermeier et al., 2011). This highly diverse South-American ecosystem complex formed by multiple physiognomies has approximately 28% left of its original vegetation cover (Rezende et al., 2018). Sadly, it is considered one of the ‘hottest of the hotspots’ (Laurance, 2009) and one of the three most vulnerable to climate change (Béllard et al., 2014). One of its most heterogeneous physiognomies is the Restinga vegetation, a mosaic of distinct coastal physiognomies dominated by herbs, shrubs and trees occurring side by side (Marques et al., 2015), and where peripheral plant communities face more extreme environment conditions than the hinterland forests (Scarano, 2002).

Despite exhibiting some floristic, functional and historical connections to other Atlantic Forest ones, at least 4% of Atlantic Forest plant species are endemic to the Restinga forest physiognomies (Marques et al., 2015). The biodiversity in these ecosystems is under serious threat as they are considered extremely vulnerable to climate change and highly exposed to deforestation and biological invasion (Zamith and Scarano, 2006). As such, species displacement (Pecl et al., 2017) and extinction (Waller et al., 2017), related to climate change, may entail irreversible consequences to the many aspects of biodiversity and ecosystem services. Indeed, services such as sediment retention, protection from sea-level rise and extreme high tide events are key to human well-being and climate change mitigation (Scarano and Ceotto, 2015).

Taxonomic diversity, i.e., the species composition and abundance at a given location and time, can be measured locally (alpha), among locations (beta) and regionally (gamma; Whittaker, 1972). Beta diversity indicates the degree of differentiation among communities and can be partitioned into two components: turnover and nestedness (Baselga, 2010). Disentangling these components allows beta diversity to be scaled up to regional levels (Socolar et al., 2016). The nestedness phenomenon occurs when the biotas of sites with less species are subsets of the biotas at richer sites, while spatial turnover denotes the replacement of some species by others due to environmental sorting and spatial constraints. Disentangling the beta diversity components represents an important tool for understanding the anthropogenic effects on the distribution of taxonomic diversity (Kraft et al., 2011).

Climate change is predicted to alter the environment and, by extension, the spatial distribution of biodiversity and ecosystem functioning. Species may locally adapt or track suitable conditions, however, given spatio-temporal restrictions to dispersal many species are expected to go locally extinct or retract their distribution leading to ecosystem degradation (Carpenter et al., 2001; Colls et al., 2009). Evidence shows that species responses to climate change may lead to biotic homogenization or heterogenization of ecological communities with detrimental effects to biodiversity (Hidasi-Neto et al., 2019; Socolar et al., 2016; Zwiener et al., 2018). Such changes directly affect the delivery of ecosystem services crucial to human well-being (Díaz et al., 2007). A way to assess these changes is by measuring functional diversity (Hooper et al., 2005), which represents the combined expression of functional traits. Functional diversity is not equally affected by all species, as the set of functional traits of each species matters differently to ecosystem processes (Mouchet et al., 2010). Combining different facets of diversity (i.e., taxonomic and functional) may represent an effective approach to estimate the effects of climate change on ecosystem structure and processes. In fact, considering the current land use changes and increasing impacts of climate disturbances in tropical ecosystems, measuring the different biodiversity levels adequately could account as an urgent task in Latin America (see Pearson et al., 2019).

Here we assessed the potential effects of climate change on woody plants of the Restinga vegetation in Brazil. Based on species checklists from local studies we generated ecological niche models and compiled functional traits for 796 woody plants to: i) predict the current and future distribution of taxonomic and functional diversities; ii) estimate the beta diversity between current and future scenarios and compare the relative contribution of turnover and nestedness; and iii) predict the distribution of functional traits indispensable to the delivery of multiple ecosystem services. Ultimately, our results contribute to the discussion of biotic homogenization and heterogenization that affect ecological communities and may jeopardize the conservation of biodiversity and human well-being.

In general, we found that climate change has the potential to critically alter the woody plant biodiversity in the Restinga vegetation by 2050. The results point to a potential taxonomic heterogenization and functional homogenization, which indicate the first stages of a sequential process of long-term biodiversity loss and biotic homogenization. The gauged increase of taxonomic beta diversity in pessimistic future scenarios accompanied by the predicted loss of species suggests subtractive taxonomic heterogenization, where the loss of few highly ubiquitous species boosts beta diversity (Socolar et al., 2016). In addition, the higher turnover in relation to nestedness indicates the replacement of some ubiquitous species by non-ubiquitous ones despite the observed overall richness reduction. The combination of higher beta diversity values and species loss supports the heterogenization hypothesis, which states that an increase in beta diversity corresponds to a decrease in the mean of distribution range sizes either through the incursion of micro-endemic species (e.g., non-ubiquitous species) into the study area or through the net contraction of species ranges (Ochoa-Ochoa et al., 2012). Nevertheless, the decrease in the number of narrow-ranged species may be a sign that some non-ubiquitous and/or endemic species are predicted to be lost, what may contribute to a process of taxonomic homogenization in the future. Regarding the functional dimension, however, we observed an ongoing process of functional homogenization whereby the mean of key traits changes and clings towards one direction. Future environmental filters will potentially constrain species functional diversity to more acclimated trait values, narrowing the functional space. The taxonomic and functional outcomes forecast by our results are alarming, considering that human impacts extend beyond the climatic factors, therefore the estimated detrimental impacts on biodiversity are likely conservative and may act in synergy with other anthropogenic impacts, potencially leading to an even worse scenario of expressive loss of biodiversity and functions (Hidasi-Neto et al., 2019; Prieto-Torres et al., 2020).

Climate change may impose particular ecological filters that constrain the occurrence of trait diversity. At a single trophic level, disturbance, if in low intensity and frequency, may increase species richness (McCabe and Gotelli, 2000), on the other hand, it may lead to species loss, as only species at a certain range of the functional traits are allowed to establish and perpetuate. As our models predicted species loss in local, regional and continental scales, climate change in the Restinga vegetation may represent an intense and growing disturbance phenomenon. The regional scale (study area), representing gamma diversity, may witness the loss of two species in the optimistic scenario and six species in the pessimistic, all with narrow distribution. The models indicated that the species Unonopsis aurantiaca may disappear in both contrasting future scenarios of the study area where it is endemic, what endorses the concerning result of our models. In the more localized scale (ecoregions), species loss reached striking levels, making up to 134 of projected lost species by eco-5 in the pessimistic scenario, which represents 19% of its current estimated woody plant species number. Moreover, the unsettling levels of species loss and displacement will potentially hinder ecosystem processes in the Restinga vegetation and, by extension, the provision of ecosystem services crucial to human well-being.

The general projected increase of wood density and decrease of maximum height (and leaf area in a couple of ecoregions) suggest that these ecosystems might face dryer and warmer environmental conditions in 2050, as woody plant species are driven towards a more conservative ecological strategy, although further studies are necessary to explore this matter. The above-mentioned homogenization of functional diversity is attributed to the decrease of two out of the three aspects of functional diversity measures, FRic and FDiv. This outcome indicates a narrower functional space with declining species abundances in more extreme trait values. Although FEve showed, instead, an overall increase in future projections, this result is rather expected, considering the significant loss of projected species richness and environmental suitability (proxy of abundance). Moreover, an increase in FEve, coupled with a decrease in FRic, has been observed in the latitudinal gradient of taxonomic diversity (Schumm et al., 2019) and after mass extinction events (Edie et al., 2018). In these cases, despite major loss of species, the majority of, if not all, functional groups will persist, even with very few species and lower abundances, so the distribution of the species' abundances tends to be more uniform.

The Restinga is one of the most vulnerable marginal ecosystems of the Atlantic Forest (Scarano, 2009). Our models have indicated drastic effects of climate change on the diversity and functionality of these systems in a near future. Nevertheless, other current menaces such as deforestation, biological invasion and land-use change (Zamith and Scarano, 2006) are not expected to decrease nor cease in the near future, and they can act in synergy with climate change. The future conservation of the Restinga can be more uncertain than the outcome pointed out in this work, as we have addressed only one threat. It is important to highlight that this vegetation is typically composed by species with high phenotypic plasticity (Zamith and Scarano, 2006), which could affect species distribution. Another relevant aspect is that part of these Restinga plant communities is also structured by the facilitation process, especially in non-forest areas at initial stages of succession (Dalotto et al., 2018). Thus, incorporating the information on the species phenotypic plasticity, biological interactions and even considering sea-level rise in future models could promote higher refinement to predictions of the effects of climate change on the Restinga vegetation.

The used method for estimating abundance from environmental suitability is a practical approach to capture changes in species distributions and their functionalities. Despite the evidence of significant correlation between the suitability index and abundance, several factors may contribute to some degree of uncertainty in the results. For instance, (i) ENMs may inaccurately estimate species-environment correlation due to limited environmental representability within accessible areas; (ii) local factors and biological interactions not anticipated by the models may limit the occurrence of species at a given site; and (iii) correlative models assume niche stability, when in fact it is dynamic. In spite of the methodological limitations and associated uncertainty, ENM endorses conjecturing over large spatial and temporal scales, and allows exploring macroecological community assembly processes (Distler et al., 2015).

Understanding the relationship between the distribution of the taxonomic and functional facets of biodiversity across spatio-temporal scales and different scenarios is crucial to guide conservation strategies that deal with the uncertainty of the future. The predicted higher future turnover associated to decreasing species richness should be considered when planning protected areas in the Restinga vegetation, otherwise it would risk losing species and functions (Tuomisto et al., 2003). Moreover, the conservation of the Restinga is also vulnerable to law subterfuges (Marques et al., 2015). Tackling this and many other issues to attain healthier ecosystems is paramount not only for the intrinsic value and maintenance of the Atlantic Forest biodiversity, but also to preserve ecosystem services essential to the prevalence of the Brazilian coastal natural wonders as well as the traditional peoples and communities that rely on its integrity.

None declared.