The Atlantic Forest is one of the main centers of diversity and endemism in the Neotropical region for ferns and lycophytes due to its high environmental heterogeneity and mountain ranges (Tryon 1972; Barrington 1993; Prado et al. 2015; Suissa and Sundue 2020). This phytogeographic domain harbors the highest diversity of these plant groups in Brazil. Of the 1,380 native fern and lycophyte species in the country, 929 occur in the Atlantic Forest, of which 413 are endemic to this biome, representing approximately 44.5% of the species present in the region (Brazil Flora Group 2022). Among the leptosporangiate ferns of the Atlantic Forest, Hymenophyllaceae stands out with seven genera and 55 recorded species. Within this family, the genus Hymenophyllum is the most species-rich, comprising 26 species distributed across six subgenera: Globosa, Hymenoglossum, Hymenophyllum, Mecodium, Myrmecostylum, and Sphaerocionium. Of these, ten species are endemic to the Atlantic Forest, including one species of the subgenus Globosa (Hymenophyllum caudiculatum); one species of subg. Hymenophyllum (H. megachilum); two species of subg. Mecodium (H. strumii and H. viridissimum), and six species of subg. Sphaerocionium (H. delicatulum, H. glaziovii, H. rufum, H. trapezoidale, H. venustum and H. vestitum) (Gonzatti and Windisch 2025).

The Atlantic Forest is among the most biodiverse biomes globally and the second-largest tropical rainforest in South America. With high species richness and endemism rates, it is considered one of the world’s 25 biodiversity hotspots (Myers et al. 2000; Rezende et al. 2018). Several key factors have contributed to the remarkable diversity and high levels of endemism in the Atlantic Forest, including: (i) climatic refugia during the Pleistocene, which acted as isolated suitability habitats during glacial periods, promoting population divergence and the emergence of new endemic lineages (Carnaval and Moritz 2008); (ii) the rugged topography formed approximately 5.6 million years ago, particularly in the Serra do Mar region, whose uplift modified the local climate and created novel environmental conditions that enhanced diversification and ecological heterogeneity (Simpson 1979; Marques et al. 2021); (iii) geographic barriers formed by major rivers, such as the Rio Doce and São Francisco, which may have promoted biological isolation and differentiation (Marques et al. 2021); and (iv) ecological gradients produced by the transition between humid forests in the central portion of the biome and adjacent drier biomes, such as the Cerrado and Caatinga (Marques et al. 2021). Together, these processes have driven the cumulative rise in biodiversity throughout the evolutionary history of the Atlantic Forest (Marques et al. 2021).

Among the factors mentioned above, climatic refugia played a fundamental role in maintaining and generating biodiversity during Quaternary climate fluctuations in the Atlantic Forest (Carnaval and Moritz 2008). During glacial periods, climate oscillations led to forest contraction, isolating populations in areas that remained environmentally stable and suitable (Carnaval and Moritz 2008). These refugia, therefore, favored genetic differentiation and the emergence of endemic lineages through geographic isolation during periods of climatic instability, thus contributing to the maintenance and accumulation of species and lineage diversity throughout the Quaternary (Carnaval et al. 2009, 2014). Apart from the Pleistocene climatic refugia, the Neotropics biodiversity has been shaped by Neogene geological events and biotic factors (Rull 2011; Álvarez-Presas et al. 2014).

Hymenophyllaceae comprises very ancient lineages, with divergences dating back more than 170 million years, and represents one of the earliest-diverging lineages among leptosporangiate ferns (Pryer et al. 2004; PPG 2016; Del Rio et al. 2017). The family underwent diversification pulses that coincided with the emergence and expansion of angiosperms, likely driven by increased substrate complexity and higher humidity promoted by flowering plants (Schneider et al. 2004; Hennequin et al. 2013). Currently, the occurrence of filmy ferns is closely associated with regions of mild temperatures and high humidity, such that species distributions remain strongly linked to climatically buffered conditions (Zotz and Büche 2000; Ye et al. 2025). This fern group is highly vulnerable to climatic variation due to a lack of efficient physiological mechanisms for water and thermal regulation (Gessner 1939; Zotz and Büche 2000; Dubuisson et al. 2003). Their leaf blades, composed of a single cell layer, lacking stomata, and with a reduced or absent cuticle, make these plants highly dependent on constantly humid, shaded environments (Iwatsuki 1990; Ebihara et al. 2007; Parra et al. 2009; Proctor 2012). Although some species exhibit a certain degree of desiccation tolerance through a poikilohydric strategy, this capacity provides only temporary protection against water loss (Dubuisson et al. 2009; Proctor 2012; Fallard et al. 2018).

In addition to their physiological peculiarities, Hymenophyllaceae exhibit reproductive traits that further underscore their strong dependence on shaded, permanently humid environments. All species in the family produce chlorophyllous spores that exhibit extremely rapid germination and short viability. This strategy contrasts with that of ferns with non-chlorophyllous spores, which exhibit greater longevity and require more extended germination periods. As a result, species with green spores are restricted to wet mesophytic habitats (Lloyd and Klekowski 1970). Therefore, Hymenophyllaceae, particularly species of Hymenophyllum, are strongly associated with high elevation and humidity, where they often reach their highest abundance and diversity at formations such as montane forests and cloud forests (Ye et al. 2025).

The endemic Hymenophyllum species of the Atlantic Forest present distinct conservation status. Hymenophyllum caudiculatum, H. megachillum, and H. sturmii are classified as Least Concern; H. rufum, H. venustum, H. vestitum, and H. viridissimum are considered Vulnerable; whereas H. delicatulum is classified Endangered (Gonzatti et al. 2020, 2023a, b; Larsen et al. 2020). Additionally, H. delicatulum is included in the list of endangered species of Rio Grande do Sul state (Decree 52.109/2014). Of the ten endemic Hymenophyllum species recognized for the Atlantic Forest, H. glaziovii and H. trapezoidale have not been assessed for conservation status. The former lacks a well-defined taxonomic delimitation, whereas the latter may be considered locally extinct, based on a single record over 70 years (Gonzatti et al. 2020).

Although climate change affects all plants, endemic species are particularly vulnerable due to their restricted habitats and dependence on specific environmental conditions, thereby increasing their risk of extinction (Manes et al. 2021; Rosete et al. 2025). In addition to limited geographic ranges, these species often exhibit specialized ecological niches, low dispersal capacity, and small population sizes, characteristics that intensify their vulnerability (Manes et al. 2021). For this reason, detailed scientific knowledge on endemic taxa, including their distribution and ecology, paired with planned conservation strategies, is essential for species persistence over time (D’Antraccoli et al. 2023). In the current context of climate change, climatic refugia continue to serve as essential areas for biodiversity persistence, as they maintain favorable environmental conditions and remain relatively buffered from climatic shifts. However, these areas are becoming increasingly vulnerable to other environmental pressures, such as land-use change (Keppel et al. 2011, 2024).

In the Atlantic Forest, intense and unregulated deforestation has resulted in severe habitat fragmentation (Ribeiro et al. 2009). This fragmentation, combined with the effects of climate change and increasing anthropogenic pressures, directly threatens the ecological integrity and survival of species that depend on stable habitats to persist over time (Ribeiro et al. 2009; Gasper et al. 2021; IPCC 2023), such as Hymenophyllum species. Thus, identifying and protecting climatic refugia is crucial for the conservation of fern species, making it essential to understand the factors and processes underlying their formation in order to prioritize areas capable of maintaining long-term stability (Keppel et al. 2011, 2024). Ecological niche modeling is therefore a fundamental tool for investigating the impacts of climate change on endemic and rare species, enabling current and future species distribution monitoring and the development of effective conservation strategies (Peterson and Vieglais 2001; Silva 2013; Qazi et al. 2022; Rosete et al. 2025).

Given the environmental heterogeneity of the Atlantic Forest and the strong association of Hymenophyllum species with humid and climatically stable environments, we expect that species diversity will be higher in mountainous regions, where environmental complexity and moisture availability are greater (Suissa et al. 2021; Kessler et al. 2011, 2016; Salazar et al. 2013). In tropical mountains, the combination of high humidity, mild temperatures, and pronounced environmental gradients promotes niche differentiation and supports high levels of fern diversity. Additionally, we expect that areas of high endemism will coincide with regions of long-term climatic stability, particularly those identified as climatic refugia, reflecting the role of stable environments in promoting persistence and diversification over time (Carnaval and Moritz 2008; Carnaval et al. 2009; Rull 2011; Keppel et al. 2024). Still, the literature lacks information on the current extent of protected areas for Atlantic Forest Hymenophyllum species, without which effective conservation policies may not be established. Moreover, similar to other species of restricted distribution (Gasper et al. 2021; Araújo et al. 2025; Angeli et al. 2026), a drastic reduction in suitability areas under protection for Hymenophyllum is likely to take place under climate change scenarios, a factor that could impact their extinction risk and further impair the prospects of conservation. Furthermore, although all species are endemic to the Atlantic Forest, we expect species-specific responses, as species with broader niche breadths will show greater resilience to climate change than those with narrower niche breadths (Slatyer et al. 2013).

This study evaluates the distribution of eight Hymenophyllum species endemic to the Atlantic Forest over the past ~ 140,000 years and under future climate change scenarios, to identify climatically stable areas through time and inform conservation strategies. Species distributions were modeled under present climatic conditions (1970–2000), past climates including the Early Holocene (HOL; 11.7–8.3 kya), the Last Glacial Maximum (LGM; ca. 21 kya), and the Last Interglacial (LIG; ca. 130 kya), as well as under future climate scenarios for 2061–2080 and 2081–2100. Specifically, we address the following questions: (i) whether Hymenophyllum species of the Atlantic Forest share a common climatic niche and exhibit similar responses to past and future climate changes, or whether responses are species-specific; (ii) investigate whether historically suitable areas coincide with current patterns of species richness and remain stable under future climate change scenarios; (iii) whether the habitat suitability area for Atlantic Forest species is placed mainly within montane environments or includes other formations (iv) whether climatically stable areas through time can be identified as priority regions for in situ conservation; (v) whether current protected areas are sufficient for the future persistence of Hymenophyllum and and how they could be improved under climate change; and (vi) to what extent projected changes in suitable climatic conditions indicate potential extinction risk under future climate scenarios.

Study area

The Atlantic Forest is located along the eastern coast of South America, with most of its extent within Brazil, and smaller portions extending southward into Argentina and Paraguay (Morellato and Haddad 2000). The biome’s topography includes mountains, plateaus, plains, and depressions, with elevations ranging from sea level to approximately 2,891 m (Ribeiro et al. 2009; Marques et al. 2021). The climate can be classified as tropical, tropical highland, and subtropical, with mean temperatures ranging from 14 °C to 21 °C and annual precipitation between 1,200 and 2,500 mm (Guedes et al. 2005). These characteristics confer highly heterogeneous environmental conditions, which are reflected in a wide diversity of physiognomies and ecosystems (Ribeiro et al. 2009; Marques et al. 2021). As a consequence of this environmental heterogeneity, the Atlantic Forest encompasses a broad range of vegetation formations (Marques et al. 2021). Native forest types include Dense Ombrophilous Forest, Mixed Ombrophilous Forest (Araucaria Forest), Open Ombrophilous Forest, Seasonal Semideciduous Forest, and Seasonal Deciduous Forest. In addition, associated ecosystems occur, including mangroves, restingas, high-altitude grasslands, inland wetlands, and northeastern forest enclaves, as defined by Law nº 11,428 of December 22, 2006.

Considering that the eight species included in this study are endemic to the Atlantic Forest biome, all analyses were conducted within a geographically defined area encompassing the biome. The following geographic limits delimited the study area: from 38.75° S to 3.71° S in latitude and from 64.08° W to 32.71° W in longitude. This extent encompasses the entire distribution of the Atlantic Forest in South America and includes an additional buffer around the biome to account for potentially suitable areas and possible range expansions. Furthermore, the delimited area corresponds to the overall distribution pattern of all species included in the study. Thus, the modeling procedures were restricted to an environmentally relevant region for the species under study. In addition, the ecological niche modeling analysis was calibrated and projected using the aforementioned geographic limits for comparison among species.

Taxa and data sampling

Recent Hymenophyllum taxonomic revisions have refined species limits and resolved previous uncertainties within the group (Gonzatti et al. 2020, 2023a, b; Larsen et al. 2020). In light of this development, our study is focused on the analysis of eight Hymenophyllum species endemic to the Atlantic Forest (Fig. 1). The distribution patterns, altitudinal ranges, growth habits, and conservation status of these species are described in Table 1. Among the ten endemic species of Hymenophyllum in the study area, only H. glaziovii and H. trapezoidale were not included in the sampling.

Occurrence records were obtained from the SpeciesLink database (https://specieslink.net/), the Reflora Virtual Herbarium (http://reflora.jbrj.gov.br/reflora/herbarioVirtual/), and the Global Biodiversity Information Facility (GBIF) platform (http://www.gbif.org). Additional data were compiled from taxonomic revisions (Gonzatti et al. 2020, 2023a, b; Larsen et al. 2020), and all specimen identifications were reviewed for accuracy. Unexamined herbarium specimens by taxonomy experts were included only if high-quality photos allowed accurate taxonomic identification. Georeferenced data was manually reviewed to ensure spatial accuracy. Records with incomplete or missing spatial information, as well as those located at city centroids, were excluded or corrected based on label information from the exsiccate, using the geoLoc online tool (https://splink.cria.org.br/geoloc). A minimum distance of 5 km was adopted between occurrence points, corresponding to the spatial resolution of the climatic variables used (2.5 arc-min ≈ 5 km). This procedure ensures that two records do not share exactly the same environmental values, thereby avoiding information redundancy and reducing spatial autocorrelation among nearby occurrences (Boria et al. 2014). When occurrence points were located within 5 km of each other, only one was retained. This spatial filtering was performed manually. The list of geographic points included for each species, along with the vouchers and other relevant information, is provided in table S1.

In total, 569 occurrence records were retained from an initial dataset of 6,633 records obtained from GBIF, SpeciesLink, Reflora, and taxonomic revisions. After taxonomic validation and verification of occurrence points, duplicate records, records lacking geographic coordinates, and those located at municipality centroids were removed. Considering each species individually, 161 records were retained for Hymenophyllum caudiculatum out of 3,349 initial records; 9 out of 101 for H. delicatulum; 46 out of 345 for H. megachilum; 50 out of 739 for H. rufum; 183 out of 1,038 for H. sturmii; 55 out of 276 for H. venustum; 43 out of 703 for H. vestitum; and 22 out of 81 for H. viridissimum. Additionally, table S2 provides a summary of occurrence records for Hymenophyllum species analyzed in this study from different data sources.

Climatic data

Habitat suitability models were built using 19 bioclimatic variables and elevation from WorldClim v2.1 (Fick and Hijmans 2017) and PaleoClim (Brown et al. 2018). Current and past climate scenarios were projected for the following periods: the present (1970 − 2000), early-Holocene (HOL; 11.7–8.3 kya) (Fordham et al. 2017), the Last Glacial Maximum (LGM; ca. 21 kya) (Karger et al. 2017); and the Last Interglacial (LIG; ca. 130 kya) (Otto-Bliesner et al. 2006). Future climate scenarios were derived from projections of the Coupled Model Intercomparison Project Phase 6 (CMIP6) datasets (O’Neill et al. 2016; Zelinka et al. 2020), which are also available at the WorldClim platform. Five Global Circulation Models (GCMs) were selected for the periods 2061–2080 and 2081–2100: CMCC-ESM2 developed by the Centro Euro-Mediterraneo sui Cambiamenti Climatici from Italy (Vichi et al. 2011; Cherchi et al. 2019), INM-CM5-0 developed by the Institute for Numerical Mathematics from the Russian Academy of Science (Volodin et al. 2017), MIROC6 developed by the Japan Agency for Marine-Earth Science and Technology (Takemura 2019), MPI-ESM1-2-HR developed by the Max Planck Institute for Meteorology (Jungclaus et al. 2019), MRI-ESM2-0 developed by Meteorological Research Institute from Japan (Yukimoto et al. 2019).

The selection of GCMs was based on their tested performance for the South American region and their availability in WorldClim (Fick and Hijmans 2017; Cannon 2020; Bazzanela et al. 2024; Reboita et al. 2024). The MPI-ESM1-2-HR, INM-CM5-0, and MRI-ESM2-0 models perform well across the Northeastern and Southeastern regions of South America, which encompass the Atlantic Forest, reinforcing its suitability for applications in this region (Bazzanela et al. 2024). As for MIROC-6, it has been increasingly recognized as one of the most accurate GCMs for Central and South America (Ortega et al. 2021; Oliveira et al. 2025). Additionally, the CMCC-ESM2 and MPI-ESM1-2-HR models have been verified as two of the best-performing models on air temperature and precipitation, respectively, for the Southeastern region of Brazil (Reboita et al. 2024), which includes most of our study area. The use of these models is supported by their widespread use in recent studies on species distribution modeling for the Atlantic Forest region and adjacent biomes (Monteiro et al. 2023; Muniz et al. 2024; Oliveira et al. 2024, 2025; Barrientos-Díaz et al. 2024; Mota et al. 2025, 2026; Passos et al. 2025; Montemayor et al. 2025; Schwantes et al. 2025).

In addition, two Shared Socioeconomic Pathways (SSPs) were considered: SSP5-8.5, a pessimistic scenario without mitigation strategies where greenhouse gas emissions rise sharply due to a rapidly growing economy dependent on fossil fuels, resulting in global warming of about 4 to 5 °C by 2100; and the optimistic SSP1-2.6 scenario, which assumes effective mitigation policies such as reducing pollutant emissions from fossil fuels, limiting warming to approximately 1.5 to 2°C by the end of the century (O’Neill et al. 2016; Tebaldi et al. 2021). All climate data (past, present, and future) were obtained at a spatial resolution of 2.5 arc-minutes (~ 5 km²).

For each future climate scenario and time period, the separate GCM ecological niche projections were combined using a consensus averaging approach. The resulting consensus maps were used in subsequent analyses. Mean-based consensus methods provide significantly more robust predictions than individual models and other consensus approaches (Marmion et al. 2009). This approach assumes that projections from different GCMs are equally plausible and helps reduce the influence of individual model uncertainties (Cooper et al. 2016; Lemoine 2021).

The climate layers were trimmed to encompass the full extent of the Atlantic Forest, following the previously delineated study area, using the raster package in R (Hijmans 2025a). A pairwise correlation analysis was conducted for the bioclimatic variables to exclude highly correlated variables and retain the most informative ones for modeling, as recommended by Soley-Guardia et al. (2024). Collinearity among present-day environmental variables was assessed using Pearson correlation coefficients. For pairs of highly correlated variables (|r| ≥ 0.7), the variable with the lowest Akaike Information Criterion (AIC) value was retained and used to create the distribution models. This step was carried out using the fuzzySim package in R (Barbosa 2015).

Thus, based on the Pearson correlation coefficient (|r| ≥ 0.7), the following variables were selected: annual mean temperature (bio1), mean diurnal range (bio2), isothermality (bio3), temperature seasonality (bio4), temperature annual range (bio7), annual precipitation (bio12), precipitation of the wettest month (bio13), precipitation of the driest month (bio14), precipitation seasonality (bio15), and precipitation of the wettest quarter (bio18). The selected variables for each species were: Hymenophyllum caudiculatum (bio1, bio2, bio3, bio12, bio13, bio14); H. delicatulum (bio1, bio4, bio12, bio15, bio18); H. megachilum (bio1, bio2, bio12, bio13); H. rufum (bio1, bio3, bio7, bio12, bio13); H. sturmii (bio1, bio2, bio12, bio13, bio14); H. venustum (bio1, bio12, bio13, bio14); H. vestitum (bio1, bio2, bio3, bio12, bio13, bio14); and H. viridissimum (bio1, bio2, bio3, bio14).

Species distribution modeling

The species distribution modeling was performed using an ensemble approach with the biomod2 package in R (Guéguen et al. 2025). Models for each species were calibrated using present-day climatic data and subsequently used to project climatic suitability for each species under paleoclimatic periods and future climate scenarios. Pseudo-absence points were generated using the “random” strategy (Barbet-Massin et al. 2012; Guéguen et al. 2025), testing three sampling alternatives: (i) number of pseudo-absences equivalent to three times the number of presences, in ten replicate sets; (ii) 1,000 pseudo-absence points in ten replicate sets; and (iii) 10,000 pseudo-absence points in three replicate sets. The goal was to obtain a distribution more consistent with the ecology and known distribution of the species, as well as greater robustness in the evaluation metrics area under the ROC curve (AUC-ROC) and true skill statistic (TSS) (Allouche et al. 2006; Phillips et al. 2006). Overall, the three sampling alternatives achieved outstanding and similar results (Table S3). The iteration with ten replicates and pseudo-absences set to three times the number of presences was selected for further analyses based on the specified criteria. The other iterations and their evaluation metrics are available in table S3.

For model validation, 80% of the occurrence records were randomly selected for training, and the remaining 20% were used for testing. Ten different algorithms were employed: Classification Tree Analysis (CTA), Flexible Discriminant Analysis (FDA), Generalized Additive Model (GAM), Generalized Boosting Model (GBM), Generalized Linear Model (GLM), Multiple Adaptive Regression Splines (MARS), Maximum Entropy (MAXNET and MAXENT), Random Forest (RF), and eXtreme Gradient Boosting (XGBOOST). Model performance was evaluated using the Area Under the ROC Curve (AUC) and the True Skill Statistic (TSS). Only individual models with a TSS of 0.7 or higher were selected. Finally, an ensemble model was built for each species to estimate the potential distribution area under different climate scenarios (past, present, and future). The contribution of each bioclimatic variable was assessed using the bm_VariablesImportance function implemented in the biomod2 package (Guéguen et al. 2025).

Niche overlap

To evaluate niche overlap across different periods and among species within the same period, we used the raster.overlap function from the ENMTools package (Warren et al. 2021). This function calculates spatial overlap indices between suitability maps, generating values for Schoener’s D and Hellinger’s I, both of which range from 0 (no overlap) to 1 (complete overlap). For intraspecific comparisons, pairwise niche overlap was estimated between current and past projections (LIG, LGM, and HOL), as well as between current and future projections under optimistic and pessimistic climate scenarios for the periods 2061–2080 and 2081–2100, for each endemic Hymenophyllum species of the Atlantic Forest. For interspecific comparisons, pairwise niche overlap was calculated among the eight Hymenophyllum species based on climatic suitability projections for each analyzed period, including current, LIG, LGM, HOL, and future optimistic and pessimistic scenarios (2081–2100).

Binarization of suitability maps

For landscape and conservation analyses, continuous suitability maps were converted into binary maps using the threshold that maximized the TSS (Das et al. 2019). The thresholds used for each species were as follows: Hymenophyllum caudiculatum (0.472), H. delicatulum (0.489), H. megachilum (0.544), H. rufum (0.537), H. sturmii (0.458), H. venustum (0.476), H. vestitum (0.497), and H. viridissimum (0.510). Values equal to or above these thresholds indicate the presence (1) of suitable climatic conditions, whereas values below indicate absence (0).

This procedure and all subsequent analyses using the habitat suitability maps were conducted in R v.4.3.1 with the packages terra v.1.8 (Hijmans 2025b), raster v.3.6 (Hijmans 2025a), sf v.1.0 (Pebesma 2018), fs v.1.6 (Hester et al. 2025), geosphere v.1.5 (Hijmans 2019), and rnaturalearth v.0.3.3 (Massicotte and South 2025).

Climatically suitable areas quantification

Climatically suitable areas for each species and time period (past, present, and future) were quantified in square kilometers (km²). Changes in suitable areas over time were assessed using the binarized maps. Current and future suitable climatic areas were overlaid with federal protected areas (excluding Private Natural Heritage Reserves), using data from the Instituto Chico Mendes de Conservação da Biodiversidade (ICMBio), updated on February 27, 2025 (https://www.gov.br/icmbio/pt-br). This spatial overlay enabled the quantification of the total extent (in km²) of climatically suitable areas within and outside protected areas, providing a direct measure of the effectiveness of protected areas under projected climate change scenarios, for both current and future conditions.

Potential species richness

The potential species richness under different climate scenarios was quantified using suitability maps generated for eight species, including the current scenario and future optimistic and pessimistic scenarios for the period 2081–2100. The future is based on the projections from the consensus of multiple GCMs. Future projections were based on the consensus of multiple GCMs. Continuous suitability maps were converted into binary maps, which were then stacked. Species richness was calculated by summing the number of species predicted per pixel, following the approach described by Massawe et al. (2025).

Climatic refugia identification

Climatic refugia were identified as regions where suitable areas for species overlap across the analyzed time periods, corresponding to the intersection of suitable areas over time (Das et al. 2019; Brasil-Godinho et al. 2024). Climate refugia maps were generated based on the overlap of the following suitability projections: (i) current refugia – paleoclimatic periods (LIG, LGM and HOL) and present; (ii) future optimistic refugia – past, present and optimistic future scenarios (2061–2081 and 2081–2100); and (iii) future pessimistic refugia – past, present and pessimistic future scenarios (2061–2081 and 2081–2100). Future scenarios were based on projections from the consensus of multiple GCMs. For these analyses, binarized maps were used. We analyzed regions of stability shared among seven of the eight species, as H. delicatulum and H. viridissimum currently exhibit distinct and non-overlapping distribution patterns, with H. delicatulum restricted to southern regions and H. viridissimum occurring in southeastern Brazil.

Extinction risk assessment

To provide a conservation-oriented interpretation of projected changes in species distributions, we assessed extinction risk under IUCN Criterion A3(c), which considers future population reduction projected, inferred, or suspected from declines in habitat quality or area (IUCN 2024). Under this criterion, species are classified according to the magnitude of the expected reduction over a specified time frame, with thresholds defined as follows: reductions ≥ 80% are categorized as Critically Endangered (CR), ≥ 50% as Endangered (EN), and ≥ 30% as Vulnerable (VU) (IUCN 2024). Estimates of the climatically suitable area (km²) used to infer these reductions were obtained from ecological niche suitability maps and are detailed in the section Climatically Suitable Areas Quantification. In addition, we estimated projected changes in suitable areas for the period 2081–2100 under both optimistic (SSP1-2.6) and pessimistic (SSP5-8.5) climate change scenarios.

Current conservation status of the eight species was obtained from the literature (Larsen et al. 2020; Gonzatti et al. 2020, 2023a, b). In these studies, extinction risk was assessed following IUCN Criterion B2, which is based on the current area of occupancy. These classifications were used here as a baseline representing present-day extinction risk, allowing comparison with future projections derived from Criterion A3(c).

To summarize projected changes in extinction risk, we used a network-based visualization approach (alluvial diagram) implemented in the R package ggalluvial (Brunson 2020). In this framework, conservation categories were treated as nodes and species transitions among categories under current, optimistic (SSP1-2.6), and pessimistic (SSP5-8.5) scenarios were represented as edges connecting nodes across scenarios.

Current distribution and habitat suitability

For the eight Hymenophyllum species, the models exhibited high predictive accuracy, with mean AUC–ROC values exceeding 0.97 and TSS values exceeding 0.88 (Table S3). Annual mean temperature (bio1) was the main contributor to niche model generation for seven species, contributing 0.526 in H. caudiculatum, 0.375 in H. delicatulum, 0.852 in H. megachilum, 0.783 in H. rufum, 0.801 in H. sturmii, 0.629 in H. venustum, and 0.848 in H. viridissimum. Diurnal temperature range (bio2) was the most influential variable for H. vestitum, accounting for 0.364 of the variance. Other precipitation variables, such as precipitation of the wettest month (bio13) and the driest month (bio14), also played important but variable roles across species (Table S4).

In general, the current habitat suitability projections match the known distribution of all eight species, with the potential area restricted to the Atlantic Forest domain (Fig. 2; Table 1). Most species showed high habitat suitability in the mountainous regions of southeastern and southern Brazil, especially along the Serra da Mantiqueira, Serra do Espinhaço, Serra do Mar, and Serra Geral ranges. Some species, such as Hymenophyllum caudiculatum and H. vestitum, showed broader and more continuous areas of suitability, whereas others, including H. megachilum, H. rufum, and H. sturmii, exhibited a more fragmented distribution pattern (Fig. 2). Hymenophyllum delicatulum presented a more restricted suitable area concentrated in southern Brazil, with a small disjunct area predicted in Bahia; however, there are currently no known records of the species in that region. Similarly, H. viridissimum, despite being currently known only from southeastern Brazil, showed additional suitable areas extending into southern Brazil.

Paleoclimatic changes in habitat suitability

Paleoclimatic niche model projections revealed both shared patterns in potential suitability areas and species-specific differences in their responses to past climatic changes (Fig. 2). In general, projections for the LIG indicated more restricted suitable areas, mainly concentrated in southern and southeastern Brazil. During the LGM, most species expanded into more suitable areas, with broader distributions than during the LIG. Seven of the eight species (Hymenophyllum caudiculatum, H. megachilum, H. rufum, H. sturmii, H. venustum, H. vestitum, and H. viridissimum) exhibited larger areas of high habitat suitability during the LGM, with expansion of suitable areas from southern and southeastern Brazil toward northeastern Brazil; in some cases, there was also broad expansion toward central Brazil, although with highly fragmented areas. The increase in area during the LGM did not always translate into spatial continuity, as evidenced by fragmentation in some species, such as H. caudiculatum and H. vestitum. In the Holocene, suitable areas were generally reduced relative to the LGM, showing a pattern closer to the present distribution, with reductions and greater fragmentation of suitable habitats. In contrast, H. delicatulum showed a distinct pattern, with greater expansion during the LIG followed by habitat contraction in later periods.

Future climatic suitability

Future climate change is projected to strongly affect the distribution of eight Hymenophyllum species, particularly in low-altitude areas outside mountainous regions (Fig. 3). This pattern was consistently observed across all GCMs used for the future projections (Fig. S1). Overall, all species are expected to lose suitable areas, especially under the pessimistic scenario (SSP5-8.5) during the 2081–2100 period. In addition, for several species, the most stable areas in future projections are concentrated along the Serra Geral, particularly in the Santa Catarina region, as observed, for example, for H. delicatulum, H. megachilum, and H. sturmii.

Hymenophyllum delicatulum and H. viridissimum show the most severe responses to future climate change. With already restricted distributions, their suitable habitat contract under all analyzed scenarios, nearly disappearing under (SSP5-8.5) by 2081–2100, and persisting under the optimistic scenario (SSP2-4.5), only as small, isolated suitable areas, indicating a high risk of habitat loss for these endemic species (Fig. 3). H. megachilum, H. rufum and H. sturmii also display a progressive reduction in climatic suitability area, mainly in the peripheral areas of their current distribution, with significant losses of continuous suitable areas and increased isolation of potential populations under the pessimist scenario (Fig. 3).

In contrast, Hymenophyllum caudiculatum, H. venustum, and H. vestitum showed a moderate response to future climate change compared to the other species. Under the SSP2-4.5 scenario, suitable areas are partially retained, with modest losses and some connectivity remaining. However, under SSP5-8.5, high-suitability areas may sharply decline and fragment by 2081–2100, risking long-term population viability (Fig. 3).

Variation in suitable habitat across time

Considering paleoclimatic periods, the quantification of suitable areas (km²) showed that, for most species, the LGM period presented the largest suitable area (Table 2). An exception was Hymenophyllum delicatulum, which exhibited its largest suitable area during the LIG period, representing a 279% increase relative to its current climatically suitable area. In general, the HOL and LIG periods showed similar extents of suitable area; however, under current conditions, most species occupy smaller suitable areas than in the paleoclimatic periods analyzed. Hymenophyllum vestitum represented an exception, showing considerable reductions in suitable area relative to the present during both the LIG (− 31%) and the LGM (− 23%).

In our future climate projections, the quantification of climatically suitable areas indicates a consistent decline. Even under conservative scenarios, a clear pattern of range contraction is observed. In the optimistic scenario, projected reductions varied among species, from relatively modest losses of about 20% for Hymenophyllum venustum and H. vestitum, to more severe losses of more than 50% for H. megachilum, H. rufum, and H. viridissimum (Table 2). For the pessimistic scenario, the projection showed even more pronounced declines for all species. Widely distributed species, such as H. caudiculatum and H. sturmii, are expected to experience substantial habitat losses of approximately 60% and up to 80%, respectively, by the end of the century. Particularly critical is the situation of species with restricted geographic ranges. Hymenophyllum viridissimum, for example, may lose up to 90% of its climatically suitable area. Similarly, H. rufum and H. megachilum are also projected to undergo severe contractions, with losses in suitable areas exceeding 90%.

Niche overlap

The assessment of intraspecific niche overlap across different climatic periods (Table S5) showed that the HOL period was most similar to the present (Schoener’s D ranged from 0.748 to 0.915). In contrast, during the LGM (Schoener’s D between 0.695 and 0.791), although suitable areas expanded, the overlap with the current distribution was lower than in other past periods, suggesting that this expansion occurred in environmentally distinct regions from those currently occupied. For the LIG (Schoener’s D between 0.640 and 0.877), niche overlap was lower than in the HOL but higher than in the LGM. For future projections, higher overlap values were observed under the optimistic scenario (SSP1-2.6) for the 2081–2100 period (Table S5). Even under this optimistic climate change scenario for the end of the century, Schoener’s D values ranged from 0.642 to 0.853.

The interspecific niche overlap analysis across different climatic periods (past, present, and future) revealed a consistently high degree of niche similarity among the analyzed Hymenophyllum species (Table S6). In the past (LIG, LGM, and HOL), species exhibited substantial overlap, which remains high in the present. Even under future climate change scenarios, both optimistic and pessimistic, the species continue to exhibit high ecological niche similarity.

Climate change impacts on protected suitable area

The extent of climatically suitable areas currently overlapping protected areas varied among species. In terms of proportional representation within protected areas, Hymenophyllum vestitum showed the highest coverage (35,275 km²; 8.81% of its total suitable area), followed by H. rufum (21,859 km²; 8.67%), H. caudiculatum (39,607 km²; 8.17%), H. megachilum (17,206 km²; 7.46%), H. sturmii (38,122 km²; 6.74%), and H. venustum (40,288 km²; 6.01%). In contrast, the narrowly distributed species H. viridissimum (18,469 km²; 3.14%) and especially H. delicatulum (3,328 km²; 2.88%) showed the lowest proportions of suitable area represented within protected areas. Overall, the proportion of protected suitable habitat remains below 10% for all species.

Projected changes in climatically suitable areas within current protected areas indicate an even stronger negative effect of climate change on species persistence (Fig. 4). Under the optimistic scenario (2061–2080), reductions within protected areas ranged from approximately 12–30% for Hymenophyllum caudiculatum, H. venustum, and H. vestitum, to more severe losses exceeding 50% for H. megachilum and H. viridissimum. By 2081–2100, contractions intensified for most species, particularly for H. megachilum, H. rufum, and H. viridissimum, which lost more than 50% of currently protected suitable areas.

Under the pessimistic scenario, projected losses were even more pronounced. By 2061–2080, reductions in protected suitable areas exceeded 75% for Hymenophyllum megachilum and H. rufum, and approached 90% for H. viridissimum. By the end of the century (2081–2100), nearly all species showed severe contractions within protected areas, with projected losses above 50% and exceeding 90% for H. megachilum, H. rufum, and H. viridissimum.

Among the analyzed species, Hymenophyllum delicatulum was the only species projected to show gains in suitable area within protected areas under some future scenarios, with increases under the optimistic late-century scenario and the pessimistic mid-century scenario, whereas all remaining species exhibited consistent declines. Overall, these results indicate that current protected areas may lose substantial climatic suitability for most endemic Hymenophyllum species under future climate change.

Identification of climatic refugia and conservation hotspots

In the current scenario, species richness shows higher concentrations in the mountainous regions of southeastern and southern Brazil (Fig. 5-1A). Under the optimistic future scenario, a reduction in the extent of these areas is observed, although regions with relatively high richness persist, mainly in southern Brazil, with some areas of high richness also remaining in the Southeast (Fig. 5-1B). Under the pessimistic future scenario, the contraction is even more pronounced, with a strong decline in species richness and a restriction of climatically suitable areas to almost exclusively southern Brazil (Fig. 5-1C).

Regarding climatic refugia, the analysis considering areas climatically stable for at least seven of the eight taxa included, indicates that historical refugia (Fig. 2A), defined by the overlap between paleoclimatic periods and the present, comprise multiple climatically stable regions for the Hymenophyllum species analyzed. Three main refugial areas stand out: (i) the southern Atlantic Forest, particularly in the mountainous regions of the states of Santa Catarina and Rio Grande do Sul; (ii) the mountainous region of the state of Rio de Janeiro; and (iii) a fragmented area along the Serra da Mantiqueira, on the border of the states of São Paulo, Minas Gerais, and Rio de Janeiro.

When future optimistic scenarios were incorporated (Fig. 5-2B), these refugial areas became more restricted, although the southern Atlantic Forest and portions of southeastern montane regions still retained climatically stable areas. Under the pessimistic scenario (Fig. 5-2C), refugia were drastically reduced. Overall, a progressive reduction in climatically stable areas is observed across current, optimistic and pessimistic projections. Individual climatic refugia were also evaluated separately for each species, and these results are presented in figure S2. Based on the overlap between climatic refugia and areas of species richness, two priority regions for the in situ conservation of Hymenophyllum species were identified (Fig. 5-3A). Cluster A (Fig. 5-3B) is located in southeastern Brazil, encompassing the states of Rio de Janeiro, São Paulo, and Minas Gerais. It includes 14 federally protected areas, such as Environmental Protection Areas, National Forests, National Parks, and Biological Reserves. The federal protected areas within Cluster A area: Área de Proteção Ambiental da Região Serrana de Petrópolis, Área de Proteção Ambiental da Serra da Mantiqueira, Área de Proteção Ambiental Bacia do Paraíba do Sul, Área de Proteção Ambiental de Guapi-Mirim, Área de Relevante Interesse Ecológico Floresta da Cicuta, Área de Proteção Ambiental de Cairuçu, Área de Proteção Ambiental da Bacia do Rio São João/Mico-Leão-Dourado, Floresta Nacional Mário Xavier, Parque Nacional da Serra da Bocaina, Parque Nacional da Serra dos Órgãos, Parque Nacional da Tijuca, Parque Nacional do Itatiaia, Reserva Biológica de Poço das Antas, and Reserva Biológica do Tinguá. This conservation core harbors occurrences of seven species: Hymenophyllum caudiculatum, H. megachilum, H. rufum, H. sturmii, H. venustum, H. vestitum, and H. viridissimum.

On the other hand, Priority Conservation Region B is located in the southern portion of the Atlantic Forest, encompassing Santa Catarina and Rio Grande do Sul states (Fig. 5–3C). It includes five federally protected areas, such as National Forests and National Parks, and harbors six species: Hymenophyllum caudiculatum, H. delicatulum, H. megachilum, H. rufum, H. sturmii, and H. vestitum. The federal protected areas within Cluster B are: Floresta Nacional de Canela, Floresta Nacional de São Francisco de Paula, Parque Nacional da Serra Geral, Parque Nacional de São Joaquim, Parque Nacional dos Aparados da Serra.

Projected extinction risk under climate change

Based on current conservation assessments derived from the literature, three species (Hymenophyllum caudiculatum, H. megachilum, and H. sturmii) are currently classified as Least Concern (LC), four (H. rufum, H. venustum, H. vestitum, and H. viridissimum) as Vulnerable (VU), and H. delicatulum as Endangered (EN) (Fig. 6). Applying IUCN criterion A3 based on projected future reductions in climatically suitable area resulted in a marked worsening of conservation status in both climate change scenarios for the period 2081–2100.

In the optimistic scenario (2081–2100), Hymenophyllum venustum and H. vestitum remained classified as VU, while H. delicatulum remained in the Endangered category. In contrast, H. megachilum and H. sturmii, currently classified as LC, shifted respectively to the EN and VU categories. In this same scenario, H. rufum and H. viridissimum, currently classified as VU, were projected to shift to EN as a consequence of suitable area losses (Fig. 6). In the pessimistic scenario (2081–2100), conservation outlooks became even more severe, with four species classified as EN (H. caudiculatum, H. delicatulum, H. venustum, and H. vestitum) and four projected as CR (H. megachilum, H. rufum, H. sturmii, and H. viridissimum) (Fig. 6). Overall, the projections indicate a generalized escalation in extinction risk for endemic Hymenophyllum species under future climate change.

Our results reveal that the distribution of endemic Hymenophyllum species in the Atlantic Forest is strongly shaped by climatic conditions associated with humid montane environments and mild temperatures. Ecological niche models indicate dynamic temporal changes in habitat suitability, with a general expansion of suitable areas during the LGM, followed by contraction toward the present and severe projected losses under future climate change scenarios, particularly under the pessimistic scenario. Despite interspecific differences in sensitivity to climate changes, all taxa showed reductions in climatically suitable areas and increased extinction risk under future climate change scenarios, especially species with restricted distributions. In addition, our analyses identified climatically stable refugial regions concentrated in mountain ranges of southern and southeastern Brazil, which coincide with areas of high species richness and define priority conservation hotspots. Taken together, these results highlight the dual role of past climatic stability in shaping current diversity patterns and of refugia as potential buffer areas for the persistence of Hymenophyllum species under ongoing climate change.

Our results support the hypothesis that the abundance of Hymenophyllaceae species is influenced by habitat diversity in mountainous and rugged terrain, especially among species of the genus Hymenophyllum, which predominate in forested environments at higher elevations (Ye et al. 2025). This restriction likely reflects the absence of evapotranspiration control mechanisms in Hymenophyllaceae, which makes these species highly dependent on humid montane environments, where evapotranspiration is reduced by constant humidity and mild temperatures, thereby ensuring their occurrence (Gehrig-Downie et al. 2012; Proctor 2012). Consequently, the group’s high diversity is associated with regions where the annual mean temperature ranges from approximately 15 °C to 23 °C and annual precipitation exceeds 1200 mm (Ye et al. 2025).

This pattern also follows a global trend in which fern richness and endemism are concentrated in montane areas of tropical and subtropical regions, primarily shaped by climatic variables such as annual precipitation and minimum temperature (Suissa et al. 2021). Indeed, the biodiversity hotspots for Neotropical ferns are predominantly located in mountainous regions, including southern and southeastern Brazil, where the Atlantic Forest serves as a primary center of endemism and species richness (Suissa and Sundue 2020). In this region, the complex geological history has shaped current patterns of diversity and distribution in fern lineages (Suissa and Sundue 2020).

Another important factor that appears to influence the geographical distribution of species, particularly the extent of their suitable niche area, is their growth habit. Species such as Hymenophyllum viridissimum, which is strictly epiphytic (Gonzatti et al. 2023b) and has a restricted potential distribution, exemplify how dependence on specific microenvironments can limit tolerance to environmental variation. Although epiphytism is a recurrent and evolutionarily advantageous strategy for occupying canopy and understory niches, it also imposes additional physiological challenges, such as greater exposure to microclimatic instability and limited water and nutrient availability (Zotz and Hietz 2001; Watkins and Cardelús 2012; Wanek and Zotz 2011), increasing the vulnerability of these species to environmental change and restricting their geographic range. In contrast, H. caudiculatum, which exhibits multiple growth habits (terrestrial, rupicolous, and epiphytic) (Larsen et al. 2020), is among the endemic species of the genus with the widest potential distribution in the Atlantic Forest, indicated by our projections. In this way, the different growth habits condition a high potential for acclimation and phenotypic plasticity to cope with environmental variability (Bazzaz 1991). This reinforces the idea that species occurring across a wide range of environments and exhibiting adaptations to different climatic conditions tend to display greater tolerance to environmental changes than specialist species with restricted geographic distributions (Hsu et al. 2014).

Ecological niche modeling projections for Hymenophyllum species endemic to the Atlantic Forest indicate a dynamic history of range expansion and contraction throughout the Pleistocene, closely linked to paleoclimatic oscillations. Overall, an expansion of climatically suitable areas was observed from the LIG to the LGM, followed by a contraction during the HOL, suggesting that glacial and interglacial cycles may have significantly shaped the distribution of these species. Given that Hymenophyllaceae are essentially forest ferns, dependent on humid, shaded environments with dense vegetation cover (Ebihara et al. 2007; Parra et al. 2009; Proctor 2012), the projected expansion during the LGM may be associated with humid forest formations in regions that became climatically favorable during that period. Palynological evidence and paleoclimatic modeling suggest that in some areas of southern and southeastern Brazil, forest vegetation persisted even during the coldest Pleistocene periods, forming a mosaic alongside grassland vegetation (Behling 2002; Behling et al. 2004). Although the LGM is generally described as a cold and dry period (Lora et al. 2023), which would limit forest formation (Morley 2011), there is evidence of a latitudinal humidity belt across the subtropical region of the Southern Hemisphere, including the Brazilian Atlantic rainforest region, between approximately 24 and 21 ka ago, which may have favored the persistence of these forest remnants (Ledru et al. 2005).

Another indication of forest cover persistence in southern and southeastern Brazil during this period is observed in the dynamics of the Araucaria Forest, which expanded during glacial events (Vasconcellos et al. 2024). In this context, the Hymenophyllum species included in this study, H. caudiculatum, H. delicatulum, H. rufum, H. sturmii, H. venustum, and H. vestitum, occur in Mixed Ombrophilous Forest formations, characterized by the presence of Araucaria angustifolia, which is a typical component of the vegetation.

Accordingly, our results point to the existence of historical climatic refugia: (i) one located in the southern Atlantic Forest, encompassing the mountainous regions of the states of Santa Catarina and Rio Grande do Sul; (ii) another in the highlands of the state of Rio de Janeiro; and (iii) a third fragmented nucleus along the Serra da Mantiqueira, at the border between the states of São Paulo, Minas Gerais, and Rio de Janeiro. These areas coincide with the centers of diversity and endemism of ferns and lycophytes identified in the Atlantic Forest by Souza et al. (2021), suggesting that historical climatic refugia played an essential role in maintaining biodiversity through time (Harrison and Noss 2017) and help explain the current patterns of species richness. This concordance reinforces the hypothesis that regions of high endemism are associated with long-term climatic stability, acting as persistence areas where diversity could be maintained even under intense environmental fluctuations. In these regions, stable microclimatic conditions likely buffered the effects of past climatic oscillations, thereby favoring the survival and diversification of lineages over time (Harrison and Noss 2017).

Moreover, the mountainous regions of the Serra do Mar, Serra Geral, Serra da Mantiqueira, and Serra do Espinhaço have been widely recognized as important historical climatic refugia for various biological groups, including both plant and animal species (Carnaval et al. 2009; Santos et al. 2023; Brasil-Godinho et al. 2024; Esser et al. 2024). The complex topography and high environmental heterogeneity of these mountain ranges favor the formation of stable regions that buffer climatic variation, creating suitable conditions for the persistence of populations during periods of climatic instability (Carnaval and Moritz 2008; Carnaval et al. 2009; Harrison and Noss 2017). These mountainous environments function as microclimatic refugia amid global warming, providing stable habitats for species sensitive to temperature and humidity (Rull 2009; Carnaval et al. 2014). Overall, the combination of climatic and geographic factors allows these regions to act as reservoirs of diversity from the Pleistocene to the present.

However, despite the presence of climatically stable areas over time that acted as refugia during Pleistocene climatic oscillations (Carnaval et al. 2009; Werneck et al. 2010; Souza et al. 2021; Vasconcellos et al. 2024), contemporary and projected future climate changes are occurring at a significantly faster pace than in the past (Tierney et al. 2020). This rapidity imposes new challenges to species persistence, especially for endemic species from regions such as South America, which harbors unique climates (Loarie et al. 2009; Urban 2015). Even though some species may have habitats that appear suitable under global warming, this does not mean that the areas they inhabit or migrate to will provide resources necessary for their establishment (Weiskopf et al. 2020; Ledig et al. 2010). Habitat degradation and landscape fragmentation may hinder population migration and persistence, even for those species projected to expand their suitable areas under extreme climate conditions (Holyoak and Heath 2015; Galetti et al. 2021; Faillace et al. 2021).

In the case of Hymenophyllum species restricted to the Atlantic Forest studied here, this scenario becomes particularly concerning, given that the projections obtained in this study indicate that, although some species may maintain their climatically suitable areas in the future, at least under optimistic scenarios, such as H. caudiculatum, H. vestitum, and H. venustum, most species tend to exhibit substantial reductions and fragmentation in their suitability areas, such as H. rufum, H. megachilum, and H. viridissimum. Habitat loss and fragmentation compromise connectivity among populations and reduce their effective size, increasing the risks of genetic drift and inbreeding. Consequently, these processes decrease genetic diversity and raise the genetic load of deleterious mutations. Although these effects may be more evident in the long term, they contribute significantly to the extinction risk of affected populations (Pinto et al. 2024).

Thus, understanding the dynamics of stability and historical climate refugia, along with future projections, can help identify key conservation areas that preserve both historical and contemporary species diversity. Past climate refugia are important for plant conservation because they represent areas where species have persisted through long-term climatic fluctuations over evolutionary timescales, often acting as reservoirs of genetic diversity and endemism (Keppel and Wardell-Johnson 2012; Brown et al. 2020). These regions, therefore, preserve not only current biodiversity but also the evolutionary history and processes that generated it. In contrast, future climate refugia identify areas expected to remain climatically suitable under ongoing climate change and are essential for short- to medium-term species persistence (Keppel et al. 2015; Morelli et al. 2016). However, focusing exclusively on future refugia may overlook historically stable areas that have played a key role in maintaining biodiversity through past climatic oscillations. Consequently, integrating both past and future refugia provides a more robust and comprehensive framework for conservation planning (Keppel and Wardell-Johnson 2012; Morelli et al. 2016).

Given the impacts of global warming on biological diversity, identifying refugia is essential for planning conservation strategies and for environmental management, as they can protect species from threats such as climate change (Keppel et al. 2015; Selwood and Zimmer 2020). We identified that the mountainous regions of southern and southeastern Atlantic Forest represent areas of current and future climatic stability for the species studied, constituting priority refugia for in situ conservation. These cores, especially those located in the south, concentrate the highest overlaps of suitable areas among species and remain stable even under pessimistic climate change scenarios.

Currently, the existing protected area provides limited coverage of these climatically stable regions. On average, fewer than 10% of climatically suitable areas for endemic Hymenophyllum species are included in protected areas, and this proportion is expected to decline sharply under projected future climate change, according to our results. This pattern is consistent with previous studies reporting that areas of high richness, endemism, and climatic suitability are often only partially included within the existing protected-area network. Similar mismatches between biodiversity patterns and conservation coverage have been documented for bryophytes, ferns, birds, terrestrial mammal, and flowering plants, where key centers of diversity and endemism fall outside formal protection (Murray-Smith et al. 2009; Silva et al. 2014; Vale et al. 2018; Souza et al. 2021; Sales and Pires 2023; Lima et al. 2025). Recent evidence from endemic mosses further supports this trend, showing that climatically suitable and endemic-rich areas are underrepresented in protected areas (Araújo et al. 2025). Together, these findings highlight important gaps in conservation planning for Atlantic Forest biodiversity under future climate scenarios.

This scenario highlights the need to review and expand the protected area network to incorporate not only the regions recently identified as climate refugia but also those historically recognized for harboring environmental stability and high biological diversity. Moreover, integrated studies involving multiple taxonomic groups are recommended to identify areas of convergence among groups, which may enhance the effectiveness of management and protection actions (Galatowitsch et al. 2009; Keppel et al. 2011). In this context, for endemic Hymenophyllum species of the Atlantic Forest, priority areas for in situ conservation should focus on mountainous regions, where complex topography acts as a buffer against the effects of climate change and functions as a potential climate refugia (Albrich et al. 2020).

The identification of priority areas for the conservation of the eight species analyzed in this study under climate change scenarios goes beyond the protection of these species in isolation. The implementation of in situ conservation strategies in these climatically stable areas may function as an umbrella approach, benefiting a broader spectrum of associated biodiversity. These climate refugia harbor not only endemic species of Hymenophyllum but also a high richness of other Hymenophyllaceae, comprising more than 40 species, as well as other ferns of the Atlantic Forest, groups that are likewise sensitive to climatic changes. In this context, these areas should be considered priorities in the planning and strengthening of protected areas, as well as in the establishment of new conservation units and the expansion or enhancement of connectivity among existing ones. The incorporation of these refugia into conservation policies may significantly contribute to maintaining the phylogenetic and ecological diversity of pteridophytes, ensuring the persistence of ancient and environmentally specialized lineages in the face of intensifying climate change.

Our projections of future extinction risk further reinforce the strong vulnerability of endemic Hymenophyllum species to ongoing climate change. Even species currently classified as Least Concern are projected to shift into threatened categories under future climate scenarios, while several currently threatened taxa may reach Endangered or Critically Endangered status by the end of the century. This pattern highlights how climate change alone may rapidly intensify extinction risk, especially for endemic species (Urban 2015). In this context, broader-scale assessments have shown that climate change-induced habitat loss represents one of the main drivers of future plant extinctions. For instance, the study by Wang et al. (2026), which included approximately one-fifth of the global diversity of vascular plants, demonstrated that 7–16% of species may face a high risk of extinction, mainly due to the disappearance of climatically suitable areas required for their persistence (Wang et al. 2026).

Revisiting the questions posed in the introduction to this study, our results indicate that Atlantic Forest Hymenophyllum species generally exhibit similar responses to past and future climatic fluctuations, suggesting a broadly conserved climatic niche among species. Suitable areas expanded during the LGM and are projected to undergo strong contraction under future climate change scenarios, with persistence concentrated mainly in southern and southeastern Brazil. Nevertheless, species-specific responses were also detected, possibly associated with ecological traits such as growth habit, although this relationship still requires further investigation. Historically climatically suitable areas largely coincide with current centers of species richness and with regions projected to remain relatively stable under future climate scenarios, reinforcing their importance as long-term climatic refugia. In addition, climatically suitable areas are predominantly associated with montane environments, especially in mountainous regions of the southern and southeastern Atlantic Forest. Our analyses also identified climatically stable areas shared among species, representing priority regions for in situ conservation and forming two major conservation cores in southern and southeastern Brazil. Despite their importance, current protected areas proved insufficient to ensure the long-term persistence of endemic Hymenophyllum species, and their effectiveness is expected to decline under future climate change. Expanding and adjusting protected-area networks to include climatically stable refugia and future suitable areas may substantially improve conservation effectiveness. Furthermore, projected reductions in climatically suitable areas indicate a considerable increase in extinction risk, with all eight species potentially qualifying as Endangered or Critically Endangered under pessimistic climate scenarios according to IUCN criteria. Overall, our results demonstrate that historical climatic stability has strongly influenced current diversity patterns, while future climatic refugia represent clear priorities for conservation planning aimed at maintaining the long-term persistence of endemic Atlantic Forest Hymenophyllum species.