Climate change affects species population dynamics (Zhao 2021) and may modify climate suitability and precipitation regimes in natural ecosystems (MMA 2018). These changes may increase the risk of large-scale extinction events (Pillet et al. 2022), as documented at different times in Earth’s history (Elewa and Abdelhady 2020). At present there is growing concern that rising global temperatures, largely associated with human activities, may trigger a new extinction event (Schmidt 2021; IPCC 2023). These environmental changes can accelerate the decline in plant health in both natural environments and agricultural systems (Pautasso et al. 2012), negatively impact the distribution of endemic species (Manes et al. 2021) and increase vulnerability to extinction (Qian and Qian 2025).

Studies linking climate change projections to plant population dynamics are important for supporting conservation strategies and for understanding how environmental changes affect biodiversity in semiarid regions (Cavalcante et al. 2020). Furthermore, investigating the potential distribution of species is an essential tool for identifying priority areas for conservation, helping to direct management efforts and reduce the limitations associated with the creation of new protected areas (Diniz-Filho et al. 2009). Applying spatial analyses with a research approach, such as those based on diversity metrics (Bullong et al. 2024), can deepen the understanding of the population dynamics of endemic species, especially those with restricted and regionalized geographic distributions (Echternacht et al. 2011; Costa et al. 2024). The use of species distribution models (SDM) has also proven effective in elucidating spatial patterns in specific regions and ecosystems (Siqueira and Durigan 2007), such as Northeastern Brazil.

The potential effects of climate change on the distribution of species have been widely investigated in regions such as Asia and Europe (Qazi et al. 2022). In contrast, in Neotropical countries, particularly Brazil, which harbors one of the largest portions of global biodiversity, including a high number of endemic species, such studies remain relatively scarce (Özkan-Tümer et al. 2026). The Brazilian Northeastern concentrates important areas of high biodiversity areas, specially in Caatinga and Atlantic Rainforest Biome (Oliveira et al. 2019; Suarez-Contento et al. 2024). In this region, arid and semi-arid areas have a gap on spatial distribution modeling studies and faces additional challenges due to the scarcity of occurrence data (Vasconcelos et al. 2024). Nevertheless, recent studies in Brazil’s Northeastern region have indicated severe habitat losses for many animal and plant species, especially within the Caatinga biome (Rodrigues et al. 2015; Simões et al. 2019; Moura et al. 2023; Suarez-Contento et al. 2024), with the most critical impacts predicted for endemic species (Simões et al. 2019; Almeida et al. 2024; Suarez-Contento et al. 2024).

The Northeastern region is home to more than 11,900 species of angiosperms, making it the third most diverse region in the country in terms of flora (Flora e Funga do Brasil 2026). This diversity is accompanied by high levels of endemism and marked heterogeneity in species distribution, particularly across biomes such as the Caatinga and the Atlantic Rainforest, especially in Protected Areas (Gomes et al. 2025). However, this region is also among the most vulnerable to climate change, with projections indicating reductions in precipitation and increases in temperature that may intensify aridification processes and accelerate desertification (PBMC 2013). In addition to climate pressures, extensive habitat loss driven by agricultural expansion has further reduced natural cover, contributing to the growing proportion of threatened species, especially those with restricted geographic ranges (Leal et al. 2005; Araújo et al. 2023).

These factors disproportionately affect endemic taxa, whose limited distributions and ecological specificity increase their susceptibility to environmental changes. Despite this, studies addressing species distribution and vulnerability under future climate scenarios remain scarce in Northeastern Brazil, particularly when compared to other regions of the country. This gap is especially critical in semi-arid environments, where data limitations constrain ecological niche modeling approaches, yet available evidence already indicates significant habitat loss for both plant and animal species in the Caatinga (Simões et al. 2019; Moura et al. 2023; Suarez-Contento et al. 2024).

Within this context, non-pseudanthial Euphorbioideae (Euphorbiaceae) constitute a relevant model group due to their high morphological diversity, elevated endemism, and heterogeneous distribution patterns, ranging from widespread (e.g., Excoecaria L. and Microstachys A. Juss.) to highly restricted taxa (e.g., Gradyana Athiê-Souza, A.L.Melo & M.F.Sales). Many species occur in environmentally contrasting habitats and exhibit restricted ranges and significant levels of threat (e.g. Microstachys heterodoxa (Müll.Arg.) Esser, Ophthalmoblapton pedunculare Müll. Arg., Sebastiania trinervia (Müll.Arg.) Baill.), making them particularly suitable for investigating biogeographic patterns and assessing vulnerability to climate change (Wurdack et al. 2005; Athiê-Souza et al. 2015).

In this study, we evaluated the potential impacts of climate change on endemic non-pseudanthial Euphorbioideae species in Northeastern Brazil by modeling their current and future climatic suitability under different climate scenarios projected until 2100. Specifically, we investigated changes in climatic suitability, spatial patterns of species richness, and the overlap between future suitable areas and protected areas, as well as the potential implications of these changes for extinction risk. By integrating species distribution models with conservation analyses, we aim to identify taxa and areas that may be most vulnerable to future environmental changes. We hypothesize that: (1) most taxa will exhibit proportional losses of climatically suitable areas under future scenarios; and (2) currently protected areas will not fully overlap with future areas of highest climatic suitability, resulting in conservation gaps in the effective protection of these species.

Study area

The Northeastern region is in Brazil’s tropical belt (1°02’ N, 18°20’ S, 34°47’ W, 48°45’ W) and comprises the states of Maranhão, Piauí, Ceará, Rio Grande do Norte, Paraíba, Pernambuco, Alagoas, Sergipe, and Bahia (Fig. 1). It is the third-largest region in the country in terms of territorial extension, with 1,561,177.80 km², corresponding to approximately 18.26% of the national territory (IBGE 2006). Its vast spatial dimension is reflected in significant physical, cultural, and climatic variations, which led to its subdivision into four subregions: the Mid-North, characterized as a transition zone between the Semiarid and the Amazon; the Sertão, located in the interior of the region; the Agreste, a transition zone between the Sertão and the Zona da Mata; and the Zona da Mata, located in the coastal strip that extends approximately 200 km inland (IBGE 2006). Due to the region’s remarkable environmental heterogeneity, areas of high rainfall are found along the coast, where annual precipitation can exceed 2,000 mm, in contrast to the semiarid Sertão, where annual rainfall is often less than 400 mm and droughts are recurrent. Temperatures also vary: while the coast has annual averages between 24 °C and 28 °C, the Sertão experiences high daytime temperatures, often above 30 °C.

In terms of vegetation cover, the Northeastern is home to formations belonging to four major biomes: the Atlantic Rainforest, restricted to small areas near the coast; the Cerrado, present in western Bahia and southern Maranhão; the Caatinga, predominant in the semiarid interior; and the Amazon, represented by the Mata dos Cocais in portions of Maranhão, Piauí, Rio Grande do Norte, and Ceará (Rizzini 1963). Each of these biomes presents distinct biological, hydrographic, climatic, and topographic characteristics, offering a wide range of ecological niches and environmental conditions that contribute to the region’s high biodiversity.

Study group

The study focused on species of the subfamily Euphorbioideae (Euphorbiaceae), specifically those belonging to the non-pseudanthial clade, which comprises representatives of the tribes Hippomaneae, Hureae, Pachystromateae, and Stomatocalyceae, as defined by Wurdack et al. (2005). This group is characterized by the absence of cyathium type pseudanthial inflorescences, presenting instead flowers typically arranged in racemes, thyrsus or panicles. Members of Euphorbioideae are known for producing toxic latex, rubber production (Guerra et al. 2021) and a diversity of bioactive compounds (Agra et al. 2008; Vaz et al. 2010; García-Saldaña 2024; Ramos et al. 2024; Silva et al. 2024; Souza et al. 2025a), supporting their recognized economic and pharmacological relevance.

In Brazil, Euphorbioideae is represented by 219 species, of which 100 are endemic and 26 are restricted to the Northeastern region, including 15 belonging to the non-pseudanthial group (Table 1) (Flora e Funga do Brasil 2026). These species are primarily distributed in the Atlantic Rainforest and Caatinga biomes and often exhibit restricted geographic ranges (Oliveira 2010; Esser 2012; Athiê-Souza et al. 2015; Flora e Funga do Brasil 2026). Additionally, a substantial proportion of these taxa is classified under some threat category (Athiê-Souza et al. 2015; Fernandez and Rosa 2018; Fernandez and Moraes 2020; Amorim and Fernandez 2021), which reinforces their relevance for modeling efforts aimed at assessing potential distribution shifts and conservation gaps. Taxonomic circumscription and species delimitation followed Oliveira (2010), Athiê-Souza et al. (2014, 2015), Pscheidt (2015), Cordeiro (2018) and Melo et al. (2020), and were complemented by data from Flora e Funga do Brasil (2026) database; tribal classification and group definition followed Wurdack et al. (2005).

Database – species occurrences

Occurrence data of the endemic non-pseudanthial Euphorbioideae species from Northeastern Brazil were compiled from specific literature (e.g., Wurdack et al. 2005; Oliveira 2010; Oliveira et al. 2013; Athiê-Souza et al. 2014, 2015; Pscheidt 2015; Cordeiro 2018), from 15 taxa were identified (14 species and one subspecies). Of these, 12 were included in the analyses, as they met the minimum requirement of occurrence records (> 5 records) (Supp. Table S1). Additional occurrence records were obtained from the online databases SpeciesLink (https://specieslink.net/), GBIF (https://www.gbif.org/) and Reflora (https://reflora.jbrj.gov.br/reflora/herbarioVirtual/), resulting in a total of 3,612 records after merging datasets. Species identification was validated based on protologues, the specialized literature cited above, and comparisons with type specimens. All records were manually curated to remove duplicates and records with identical geographic coordinates.

Environmental variables

Bioclimatic variables were obtained from WorldClim v2.0 (Fick and Hijmans 2017) with a spatial resolution of 30 arcseconds (~ 1 km²). All layers were cropped to the geographic extent of the Northeast region of Brazil (IBGE 2025). A correlation matrix of the climate data was calculated using Pearson’s correlation coefficients, applying a cutoff value of 70% to reduce collinearity problems. From each group of highly correlated variables, a single predictor was retained based on ecological relevance. The final set of predictors included: bio1 (Average Annual Temperature), bio4 (Temperature Seasonality), bio6 (Minimum Temperature of the Coldest Month), bio12 (Annual Precipitation), bio15 (Precipitation Seasonality), and bio18 (Precipitation of Warmest Quarter). These variables were selected to capture temperature and precipitation gradients relevant to species distribution across different biomes (e.g., Caatinga, Cerrado, and Atlantic Rainforest). A complete description of all 19 original variables, including definitions and units, is provided in Supplementary Table S1.

Species distribution modeling

Species distribution models (SDM) were developed using the Maxent algorithm (Phillips et al. 2017) through the ENMwizard package (Heming et al. 2019), as this algorithm performs well even with limited occurrence data (Elith et al. 2006; Hernandez et al. 2006; Wisz et al. 2008; Suarez-Contento et al. 2024; Lombo-Sanchez et al. 2025). The cleaned dataset was spatially filtered to reduce sampling bias and spatial autocorrelation using a distance-based thinning approach implemented in the spThin R package (Aiello-Lammens et al. 2015), applying a minimum distance of 1 km between occurrence records. This threshold was selected to match the spatial resolution of the environmental predictors (~ 1 km). After this filtering process, a total of 464 unique occurrence records were retained for subsequent analyses (Supp. Table S1). The ENMwizard package (Heming et al. 2019) was used to define the calibration area for each species, creating a minimum convex polygon around all occurrences and expanding it with a 1.5° buffer. This approach represents the areas potentially accessible to the species and increases environmental variability, improving niche predictions (Anderson and Raza 2010; Barve et al. 2011; Boria et al. 2014, 2016; Mota et al. 2022).

Model cross-validation was performed using the «ENMevaluate_b» function, and the ENMeval package was used to optimize MaxEnt parameters (Muscarella et al. 2014), allowing the evaluation of different combinations of Feature Classes (FCs) and Regularization Multipliers (RMs). Feature classes included L (linear), P (product), Q (quadratic), H (hinge), and their combinations (e.g., LP, LQH), with multipliers ranging from 0.5 to 5.0. Background points were generated within the calibration area following a presence-background approach (Heming et al. 2019).

Candidate models were evaluated using the «calib_mdl_b» function, following an Ensemble of Best-Performing Models (EBPM) approach (Heming et al. 2019). Model selection was based on statistical significance (partial ROC), omission rates (OR), and corrected Akaike Information Criterion (AICc) (Boria et al. 2016; Cobos et al. 2019; Costa-Pinto et al. 2024; Sanchez et al. 2025).

Model evaluation was conducted using a jackknife (n–1) cross-validation approach, which is particularly suitable for datasets with limited occurrence records (Veloz 2009; Hijmans 2012; Sanchez et al. 2025). Final models were generated using the «proj_mdl_b» function, producing suitability maps in cloglog format, with values ranging from 0 to 1, representing relative climatic suitability. To improve robustness, a consensus model was generated by averaging the top 10% of selected models based on the same evaluation criteria (Boria et al. 2016; Cobos et al. 2019).

The current consensus model was used to estimate changes in species distribution under future climate scenarios. Future projections (2080–2100) were generated using three general circulation models (GCMs): BCC-CSM2-MR (Wu et al. 2021), CMCC-ESM2 (Cherchi et al. 2019) and HadGEM3-GC31-LL (Williams et al. 2018). These models were selected because they cover a wide range of equilibrium climate sensitivity (ECS 2.9–5.5 °C; Zelinka et al. 2020) and have shown good performance in reproducing climate patterns in South America (Eyring et al. 2016; Zelinka et al. 2020; Hodnebrog et al. 2022; Scafetta 2023; Suarez-Contento et al. 2024; Lombo-Sanchez et al. 2025).

Climate projections were performed under two contrasting Shared Socioeconomic Pathways (SSPs): SSP126 (optimistic scenario), representing a low-emissions mitigation scenario, and SSP585 (pessimistic scenario), representing a high-emissions scenario. These scenarios were selected to evaluate species responses under opposite future climate trajectories and to compare potential biodiversity outcomes under contrasting levels of climate change severity. The study area was delimited to Northeastern Brazil, with an additional 1.5° buffer. A consensus across GCMs was applied to represent future climatic conditions and reduce uncertainty in projections.

The contraction of the species distribution area or the loss of suitable areas between current and future climatic conditions was calculated using the «thrshld_b» function of the «ENMwizard» package (Heming et al. 2019). We used the 10th percentile training presence criterion (10. percentile.training. presence), which is a method commonly used in ecological niche modelling to define potentially suitable areas (Heming et al. 2019). This method is based on the 10th percentile of presences observed during model training, thus establishing a limit that excludes the 10% of the least suitable places of presence (Boria et al. 2016; Heming et al. 2019). Suitable areas were estimated by summing the number of climatically suitable pixels in the binary models and converting them to km² according to raster resolution. Proportional changes in suitable areas were calculated as the percentage difference between current and future suitable areas for each species. Differences in the extent of suitable areas among climate scenarios were evaluated using Friedman nonparametric tests, considering taxa as repeated measures. This approach was adopted because the data did not meet normality assumptions according to the Shapiro–Wilk test. When significant differences were detected, pairwise comparisons between scenarios were performed using paired Wilcoxon signed-rank tests with Bonferroni correction to account for multiple comparisons (Kembel et al. 2010; Crawley 2013).

Potential species richness

Species richness maps were generated by stacking and summing the binary distribution models of all species for each climate scenario using the «common_spatial_metric» function of the ENMwizard package (Heming et al. 2019). Thus, each grid cell represents the predicted number of species under current and future climatic conditions (Heming et al. 2019). Spatial patterns of change were calculated in the terra package (Hijmans et al. 2020) from the difference between future species-richness rasters (SSP126 and SSP585) and the current raster, allowing the identification of areas exhibiting reduction or stability under future climate scenarios. The maps also included the spatial delimitation of protected areas to facilitate the visual interpretation of the observed patterns.

Suitable areas and protected areas

The percentage of climatically suitable areas covered by PAs was estimated for each climate scenario (SSP126 and SSP585) using the «ENMwizard» package (Heming et al. 2019). Data on protected areas were obtained from the UNEP-WCMC and IUCN databases (UNEP-WCMC and ShareAction 2024 and IUCN 2024). Differences in the extent of suitable areas within protected areas among climate scenarios were evaluated using Friedman nonparametric tests, considering taxa as repeated measures. This approach was adopted because the data did not meet normality assumptions according to the Shapiro–Wilk test. When significant differences were detected, pairwise comparisons between scenarios were performed using paired Wilcoxon signed-rank tests with Bonferroni correction to account for multiple comparisons (Kembel et al. 2010; Crawley 2013).

Estimated extinction risk

For the current conservation status, the area of ​​occupancy (AOO) of the species was used as follows: CR < 100 km², EN < 5,000 km², VU < 20,000 km², NT < 30,000 km², and LC > 30,000 km² (IUCN 2024). Conversely, for future scenarios, the conservation status analysis was performed based on the projected extent of suitable habitat derived from SDM outputs, according to criterion A3(c) (IUCN 2024). Criterion A3 considers the “projected or estimated future population reduction”, and subcriterion (c) considers the decline in area of ​​occupancy (AOO), extent of occurrence (EOO), and/or habitat quality. Therefore, species were classified according to the percentage of habitat loss following the thresholds proposed by Pomoim et al. (2022), which were based on the IUCN criterion A3(c): Extinct (EX) – 100%; Critically Endangered (CR) – above 80%; Endangered (EN) – between 50 and 80%; Vulnerable (VU) – between 30 and 50%; Near Threatened (NT) – slightly below 30%; and Least Concern (LC) – species with low or no projected habitat loss.

All analyses were carried out in the R environment using appropriate statistical, spatial, and ecological packages to process data, perform SDM and generate graphical outputs (v. 4.4.0, R Core Team 2026).

Species distribution modeling

The model’s performance was consistently high across all evaluated species. AUC values ​​ranged from 0.76 to 0.97, indicating strong discrimination power and overall predictive accuracy (Supp. Table S3). Three-quarters of taxa exhibited pronounced reductions in climatically suitable areas, under both projected scenarios, which are distributed in different biomes (Atlantic Rainforest, Caatinga and Cerrado), although the magnitude of change varied among taxa (Figs. 2, 3 and 4). Sebastiania macrocarpa and A. bahiensis were the only species that showed expansion in climatically suitable areas (Fig. 4A), increasing by approximately 39–45% under both future scenarios, whereas M. heterodoxa exhibited an increase about 33% only under optimistic scenario (Fig. 4B). Algenornia bahiensis are restricted from Atlantic Rainforest areas, while S. macrocarpa and M. heterodoxa are also distributed Caatinga.

The optimistic scenario projected reductions in climatic suitability for three-quarter of the taxa, with losses reaching 62.99% in O. pedunculare and exceeding 90% in S. trapezoidea, M. uleana, S. jacobinensis, and M. fistulifera subsp. bahiensis (Fig. 4B; Supp. Tables S4, S5 and S6). The pessimistic scenario was even more severe, A. appendiculatus showed a reduction of 54.34%, while S. jacobinensis, O. pedunculare, M. uleana, S. trapezoidea and Mabea fistulifera subsp. bahiensis were projected to lose nearly all climatically suitable areas (> 99%) (Fig. 4C).

The environmental suitability analysis revealed significant differences among climate scenarios (Current, optimistic, and pessimistic) in both the total extent of suitable areas (Friedman test: x² = 8.98, df = 2, p = 0.011) and the extent of environmentally suitable areas overlapping protected areas (Friedman test: x² = 10.26, df = 2, p = 0.006). Overall, most taxa showed low levels of protection (Fig. 5, Supp. Table S5). Under the current scenario, all taxa had less than 20% of their suitable areas located within protected areas. In future scenarios, however, some species projected to experience reductions in suitable areas showed an increase in the proportion of their remaining distribution overlapping protected areas, exceeding 20% in some cases. Taxa such as Mabea fistulifera subsp. bahiensis O. pedunculare, M. uleana, S. jacobinensis and S. trapezoidea exhibited relatively high percentages of protection under certain scenarios, although their potential distribution areas remained comparatively small. This pattern suggests that, despite their restricted distributions, a considerable portion of their projected suitable habitats may persist within protected areas, potentially favoring their long-term conservation. In contrast, more widely distributed species such as S. macrocarpa, A. bahiensis, M. heterodoxa maintained large areas of climatic suitability across scenarios, but only a small proportion of these overlapped protected areas.

Potential species richness

In general, the analyzed taxa exhibit higher richness values in Central Caatinga, Chapada Diamantina, and the Atlantic Rainforest areas along the eastern coast of Northeastern Brazil. Although these regions are projected to experience reductions under both future scenarios, they still concentrate highest levels of richness (Fig. 6). Under the current scenario, areas of highest species richness are concentrated primarily in southern Bahia, Chapada Diamantina, Araripe (located in the intersection of Ceará, Pernambuco and Piauí states) and many areas from the Pernambuco Endemism Centre (Alagoas, Paraíba, Pernambuco and Rio Grande do Norte coastal forests), reaching values of up to five species per pixel. Under the optimistic scenario, a general reduction in species richness is observed. Although some high-elevation areas in Bahia, Araripe, Pernambuco Endemism Centre, and Montane Forest Enclave of Triunfo in Pernambuco state maintain suitability conditions, the extent of zones with high richness values (≥ 4 species per pixel) decreases. Nevertheless, portions of southern Bahia and the montane regions of Chapada Diamantina, Araripe and Pernambuco Endemism Centre retained relatively high richness values, suggesting the persistence of climatically stable areas that may function as potential refugia for endemic taxa. Under the pessimistic scenario, the effects of climate change become more pronounced. Species richness declines drastically, and areas that previously exhibited high suitability shift to low or intermediate richness values (1–3 species per pixel). Only small regions in southern Bahia and Pernambuco retain moderate levels of richness. These results indicate a trend toward substantial losses of climatic suitability for the studied species, with potential consequences for regional biodiversity conservation, especially under more extreme climate scenarios.

Additional spatial analyses of richness change revealed heterogeneous responses across Northeastern Brazil (Fig. S1). Most regions exhibited stable richness values or moderate reductions under future climate scenarios, whereas localized areas of greater richness loss were concentrated mainly in coastal Atlantic Rainforest remnants and semiarid transitional regions. In contrast, portions of southern Bahia, Chapada Diamantina, and humid highland areas of Pernambuco retained relatively stable richness patterns across scenarios. The spatial overlap between richness stability and protected areas showed that some of these climatically stable regions coincide with existing conservation areas, particularly in southern Bahia and parts of the Atlantic Rainforest biome. However, several regions maintaining moderate richness values under future scenarios were located outside protected areas or occurred in fragmented landscapes.

Estimated extinction risk

The conservation status of the 12 endemic taxa from Northeastern Brazil revealed a general trend toward increasing extinction risk under future climate scenarios (Fig. 7, Supp. Table S7). Under current conditions, all taxa were classified as Least Concern. However, future projections indicate substantial changes in threat status, particularly under the pessimistic scenario. In the optimistic scenario, half of the taxa remained within the categories Least Concern and Near Threatened, with three species assigned to each category. The remaining taxa shifted to higher threat categories, including Vulnerable, Endangered, Critically Endangered and even Extinct. Under the pessimistic scenario, approximately 66,7% of the taxa were classified within the elevated extinction-risk categories. Overall, future climate scenarios resulted in a marked escalation of extinction risk across taxa, with a progressive reduction in the number of species classified as Least Concern and a corresponding increase in threatened categories, particularly under pessimistic scenario (Fig. 7). The increase in extinction risk was particularly pronounced for M. fistulifera subsp. bahiensis, which was classified as Extinct in projections for both future scenarios. In addition, M. uleana, S. jacobinensis and S. trapezoidea shifted from Least Concern under current conditions to Critically Endangered in both future scenarios. Similarly, A. appendiculatus, O. pedunculare, S. sceleratum and S. loranthacea showed progressive increases in extinction risk across scenarios. In contrast, A. bahiensis and S. macrocarpa were the least affected taxa, maintaining their current classification as Least Concern under both future climate scenarios.

Our results reveal a consistent pattern of climate vulnerability among endemic non-pseudanthial Euphorbioideae species from Northeastern Brazil. The most prominent finding was the widespread contraction of climatically suitable areas across most taxa (75%) under both future climate scenarios, particularly under pessimistic scenario, where several species were projected to lose nearly all suitable habitats. Additional key results included the concentration of future richness in a few climatically stable regions, the generally low overlap between suitable areas and protected areas, and the substantial increase in projected extinction risk for most taxa. Together, these findings indicate that climate change may severely affect the long-term persistence of endemic non-pseudanthial Euphorbioideae in Northeastern Brazil, especially species with restricted distributions associated with seasonally dry environments and montane ecosystems.

The generalized reduction in climatically suitable areas observed agrees with previous studies showing that endemic plant species are especially sensitive to climate change because they often occupy fragmented habitats and narrow environmental niches, limiting their capacity to colonize newly suitable areas (Suarez-Contento et al. 2024). This pattern may be particularly severe in seasonally dry tropical forests such as the Caatinga, where many species already occur close to their physiological tolerance limits regarding temperature and water availability (Allen et al. 2017; Silva et al. 2019; Moura et al. 2023). Increasing temperatures and changes in precipitation regimes are therefore expected to intensify water stress and reduce habitat suitability for many endemic taxa in Northeastern Brazil (Marengo et al. 2017; Andrade et al. 2017).

Although most species exhibited reductions in suitable habitats, a few taxa showed projected expansions or comparatively lower losses under future scenarios, a pattern also observed by Leão et al. (2021). These contrasting increases suggest that some species may possess greater ecological plasticity or morphophysiological adaptations associated with drought tolerance and environmental seasonality (Bongers et al. 2017; Halali et al. 2024). Furthermore, the expansion of ranges observed in some species may be associated with the maintenance of more humid climatic conditions in certain regions of the Caatinga. Peripheral areas of the domain, especially those under the influence of Atlantic-equatorial air masses and with greater water availability, tend to show less intensification of aridity in future scenarios (Moro et al. 2015; Andrade et al. 2017). The three species which presented expansion on suitable areas sharing vegetation types such as Seasonal Semidecidual Forest and Ecothons of Atlantic Rainforest overlapped with Caatinga. The projected expansion should be interpreted cautiously, since it may reflect shifts in the distribution of specific phytophysiognomies rather than a genuine reduction in vulnerability. According to the MMA (2018), Open Ombrophilous and Ecothons formations may expand under future climate conditions, potentially favoring species associated with these environments.

Similarly, the relatively moderate reduction projected for S. sceleratum may be associated with its occurrence on rocky outcrops and seasonally dry environments in the Caatinga (Cordeiro et al. 2018). Such habitats may function as ecological microrefugia under increasing aridity. Souza et al. (2025b), demonstrated that areas surrounding inselbergs in the Caatinga can maintain localized humid conditions due to runoff concentration along rocky slopes, potentially buffering the effects of regional drying. In addition, the presence of abundant laticifers, a common characteristic of Euphorbiaceae and already documented for Hippomaneae genera such as Sapium, may contribute to rapid sealing of injured tissues, thereby reduced water loss under hot and dry conditions (Demarco et al. 2013).

Likewise, the comparatively lower vulnerability of A. appendiculatus may be related to its broader geographic distribution and occurrence across distinct forest formations in the Atlantic Rainforest, since widespread species generally encounter suitable conditions across larger environmental gradients and tend to be less sensitive to climate shifts (see Leão et al. 2021). In contrast, the strongest negative impacts were concentrated in taxa associated with Caatinga vegetation and highland ecosystems, especially under pessimistic scenarios. Species such as O. pedunculare, S. trapezoidea, M. uleana, S. jacobinensis and Mabea fistulifera subsp. bahiensis exhibited severe reductions in climatically suitable areas, in some cases exceeding 99% of projected habitat loss. Similar patterns have been reported for other endemic plants from Northeastern Brazil, reinforcing the high sensitivity of seasonally dry tropical floras to future climatic changes (Suarez-Contento et al. 2024; Almeida et al. 2024; Simões et al. 2019).

Many of the most vulnerable taxa occur in Chapada Diamantina (Bahia state), a region characterized by rocky fields, elevational heterogeneity and high levels of endemism (Harley 1995). Species such as M. revoluta, S. loranthacea and M. uleana are strongly associated with these montane environments (Esser 1998; Athiê-Souza et al. 2014; Pscheidt 2015). Our projections also highlight the Pernambuco Endemism Center, an area with conservation implications essentially because half of the mammal’s species could lose suitable areas with climatic changes (Oliveira et al. 2023). In the optimistic scenario, the Montane Forest Enclave of Triunfo presents a particular climatic refuge in the mid Caatinga regions up the São Francisco River. That area is characterized by orographic conditions, enabling it as a possible climatic refuge (Ranulpho et al. 2014). Although high-altitude ecosystems frequently harbor elevated biodiversity because of their environmental heterogeneity (Chang et al. 2023), they are also particularly vulnerable to climate change due to their restricted extent and geographic isolation (Bugado et al. 2025). Previous studies have projected substantial future reductions in suitable areas for Brazilian rocky grasslands (Bitencourt et al. 2016), while ongoing deforestation and agricultural expansion continue to intensify landscape degradation in Chapada Diamantina (Funch et al. 2005, 2008; Silva et al. 2023).

Consequently, climate change may act synergistically with habitat loss and fragmentation, further compromising the persistence of endemic montane taxa. Beyond individual species responses, our results also revealed clear spatial patterns in the future distribution of species richness. Southern Bahia, Chapada Diamantina, and Pernambuco Endemism Centre consistently retained the highest richness values across scenarios, despite an overall reduction in the extent of highly suitable areas. Those regions therefore emerge as potential climatic refugia for endemic non-pseudanthial Euphorbioideae under future climate conditions. Climate refugia are increasingly recognized as essential for biodiversity conservation because they may maintain relatively stable environmental conditions over time and buffer the impacts of climate change (Morelli et al. 2020; Rossetto and Kooyman 2021). However, the persistence of richness hotspots under future scenarios does not necessarily guarantee effective long-term conservation (Kocsis et al. 2021), particularly because several climatically stable regions identified here occur outside protected areas or within highly fragmented landscapes. Importantly, the spatial coincidence between richness stability, projected habitat persistence, and limited protection coverage highlights these regions as immediate conservation priorities. In particular, southern Bahia and portions of Chapada Diamantina and Pernambuco Endemism Centre represent strategic areas where the establishment of new conservation units and habitat connectivity measures could maximize the protection of both species’ richness and future climatic refugia.

Indeed, one of the most concerning results of this study was the generally low overlap between climatically suitable areas and protected areas across all scenarios. Even under current conditions, all taxa showed less than 20% of their suitable habitats within protected areas. Although some narrowly distributed species exhibited proportionally higher overlap with protected areas under future scenarios, this pattern mainly reflected the drastic contraction of their remaining suitable habitats rather than improved protection effectiveness. Conversely, more widespread species such as A. bahiensis, S. macrocarpa, and M. heterodoxa maintained relatively large suitable areas across scenarios, but most of these regions remained outside conservation units. Similar mismatches between future suitable habitats and protected areas have been documented in other regions and may substantially reduce the effectiveness of current conservation networks under ongoing climate change (Dobrowski et al. 2021; Lombo-Sanchez et al. 2025).

The projected increase in extinction risk further reinforces the conservation concern associated with these taxa. While all species were classified as Least Concern (LC) under current conditions, future scenarios resulted in a marked escalation of threat categories, particularly under pessimistic scenario, where approximately two-thirds of the taxa shifted to elevated extinction-risk categories. Mabea fistulifera subsp. bahiensis represented the most critical case, as Extinct (EX) under both future scenarios due to the near-total loss of climatically suitable areas, suggesting that climate change may eliminate nearly all environmentally suitable conditions required for its persistence. Likewise, M. uleana, S. jacobinensis, and S. trapezoidea shifted from Least Concern to Critically Endangered under future projections. Stillingia loranthacea also deserves special attention due to its extremely restricted distribution, being known from a single municipality, combined with the projected reduction in climatically suitable areas under both future scenarios. Consequently, its conservation status shifted from Least Concern to Vulnerable. These results agree with broader evidence indicating that narrowly distributed tropical plants are disproportionately vulnerable to climate change because of their restricted environmental tolerance, small population sizes, and limited dispersal capacity (Manes et al. 2021; Silva et al. 2019). According to a global study conducted in biodiversity hotspots, high habitat specificity represents one of the main factors related to the risk of species extinction (Malcolm 2006). In addition to environmental constraints associated with habitat, the disruption of essential ecological interactions, such as pollination and seed dispersal, can further compromise the survival of endemic plant species, especially those with highly specialized reproductive strategies (Girão 2007). Conservation planning should therefore prioritize narrowly distributed taxa projected to experience severe habitat loss, especially M. fistulifera subsp. bahiensis, M. uleana, S. jacobinensis, S. loranthacea and S. trapezoidea, whose future persistence may depend on urgent in situ conservation measures, habitat protection, and long-term population monitoring.

Species known exclusively from type material (O. parviflorum, S. trinervia, and G. franciscana) were not included in the distribution models due to insufficient occurrence records. Nevertheless, these taxa are likely highly vulnerable due to their microendemism and extremely restricted populations, factors strongly associated with increased extinction risk (Qian and Qian 2025). In particular, G. franciscana, recorded from the margins of the Lower São Francisco River, a region heavily affected by human activity and intense modifications of riparian landscape, resulting in substantial impacts on associated ecosystems (Araújo 2015). Overall, our findings indicate that current protected areas may be insufficient to ensure the long-term persistence of endemic non-pseudanthial Euphorbioideae under future climate change. Conservation planning should therefore incorporate future climatic suitability into the design and expansion of protected-area networks, especially in regions identified here as potential climatic refugia, such as Southern Bahia, Chapada Diamantina, and Pernambuco Endemism Centre. In addition, conservation strategies should extend beyond currently protected areas and include landscape connectivity, habitat restoration, and long-term monitoring of vulnerable populations.

While the patterns identified here are consistent, it is important to acknowledge the inherent uncertainties associated with ecological niche modelling. Correlative models may not fully account for biotic interactions, dispersal limitations, historical processes, land-use changes, or fine-scale microenvironmental variables that can also influence species distributions. Future projections are additionally subject to uncertainties related to the performance of GCM and greenhouse gas emission scenarios (Eyring et al. 2016; Scafetta 2023). In addition, the use of a single modelling algorithm may also contribute to uncertainty in projected distributions, although MaxEnt is widely recognized as a robust approach for modelling species distributions using presence-only data, particularly for rare and endemic taxa (Elith et al. 2006; Hernandez et al. 2006; Wisz et al. 2008; Suarez-Contento et al. 2024; Lombo-Sanchez et al. 2025). Furthermore, some endemic species included in this study are represented by a limited number of occurrence records, which may affect model precision despite the use of modelling approaches specifically designed for sample sizes. Nevertheless, the consistency of the observed trends across different climate models and scenarios, together with the ecological coherence of the results, supports the reliability of the general patterns identified in this study. Finally, future studies integrating multiple modeling algorithms, dispersal data, and population genetic data could improve predictions of species responses to climate change.

Authors

1. Tiago Oliveira
2. Yeison Jaroc Lombo-Sanchez
3. Karen Yuliana Suarez-Contento
4. Alícia Marques Torres
5. Jone Clebson Ribeiro Mendes
6. Vitória Raquel da Silva Lima
7. Sarah Maria Athiê-Souza

Cite this article

Oliveira, T., Lombo-Sanchez, Y.J., Suarez-Contento, K.Y. et al. Projections of suitable habitat loss and its implications in conservation for endemic non-pseudanthial Euphorbioideae (Euphorbiaceae) species in Northeastern Brazil under climate change scenarios. Biodivers Conserv 35, 210 (2026). https://doi.org/10.1007/s10531-026-03400-1

Download citation

• Received: 04 October 2025

• Revised: 16 May 2026

• Accepted: 16 June 2026

• Published: 08 July 2026

• Version of record: 08 July 2026

• DOI: https://doi.org/10.1007/s10531-026-03400-1

Keywords