Earth is facing a global biodiversity and ecosystem crisis, with over 28% of assessed species currently threatened with extinction (Jaureguiberry et al. 2022; IUCN 2023). This decline is primarily driven by human activities, particularly changes in land use and exploitation, which have historically been identified as the most pervasive threats to biodiversity (Jaureguiberry et al. 2022) and frequently result in habitat loss and fragmentation. While these pressures remain dominant, especially in tropical regions, accelerated climate change driven by greenhouse gas emissions is emerging as an additional and increasingly pervasive driver of biodiversity loss across spatial and taxonomic scales (Scheffers et al. 2016; Jaureguiberry et al. 2022).

The global temperature increase observed over the past 75 years is unprecedented in recent millennia (IPCC 2023). Climate change alters the spatial distribution of environmental conditions suitable for species persistence, potentially leading to range shifts, contractions, or local extinctions when dispersal is limited, and affects resource availability and population dynamics through changes in plant distribution and phenology (Morellato et al. 2016; Pecl et al. 2017; Waller et al. 2017; Sales et al. 2020). Such changes can disrupt ecological interactions and ecosystem functioning, including consumer–resource dynamics among taxa with shared evolutionary histories, such as Neotropical primates and angiosperms, given the key role of primates as seed dispersers and their strong dependence on plant resources (Fontúrbel et al. 2021; Fuzessy et al. 2022). Understanding how climate change may reshape these interactions is critical for anticipating future biodiversity responses and informing conservation strategies.

Neotropical primates are predominantly arboreal or semi-arboreal and generally depend on ecosystems that provide sufficient resource availability, plant diversity, and complex ecological interactions to meet their ecological requirements (Gouveia et al. 2014). This close association with forest structure also underpins the functional importance of primates in these ecosystems. Primates are key ecological agents in tropical forests, shaping plant community structure through seed predation, herbivory, and particularly seed dispersal via frugivory, which directly affects plant recruitment, gene flow, and forest dynamics (e.g., Mourthé et al. 2008; Brocardo et al. 2010; Bufalo et al. 2016).

Approximately 65% of primate species are threatened with extinction, and ~ 93% have declining population trends, primarily due to habitat loss and fragmentation, with climate change acting as an additional and intensifying driver (Fernández et al. 2022; IUCN 2022a). Climate change is projected to increase global average temperatures by 2 °C by 2100, exposing many primate species to novel thermal conditions within their current ranges (Graham et al. 2016). As temperature and precipitation regimes shift, the spatial distribution of areas with climatic conditions suitable for species persistence is projected to contract and/or shift geographically. For Atlantic Forest primates, 74% of species are predicted to lose more than 50% of their climatically suitable areas, and 47% may lose more than 75% under future scenarios (Pinto et al. 2023). Such reductions are likely to create spatial mismatches between current forest remnants and areas projected to become climatically suitable in future scenarios. In landscapes already characterized by severe fragmentation and low forest cover, such as the Atlantic Forest, dispersal between suitable areas becomes increasingly constrained. Consequently, limited connectivity may prevent primates from tracking shifting climatic suitability successfully, thus increasing the likelihood of population declines or local extinctions.

The Atlantic Forest is a major biodiversity hotspot (Myers et al. 2000) with high projected climate change exposure (Bellard et al. 2014). Muriquis (B. arachnoides and B. hypoxanthus) are the largest extant non-human primates in the Americas, weighing on average 9.6 (males) and 8.3 kg (females), with population densities of ~ 2–50 ind./km², a multi-male and multi-female social structure, male philopatry, and dispersing females (Almeida-Silva et al. 2005; Strier et al. 2015). They are endemic to the Atlantic Forest and are both classified as Critically Endangered (Talebi et al. 2021; de Melo et al. 2021). Under future greenhouse gas emission scenarios, projections of climatically suitable areas indicate range losses of 24–43% for B. hypoxanthus and 11–39% for B. arachnoides by 2050, and 44–92% and 13–69%, respectively, by 2100 (Portillo 2021; Pinto et al. 2023; Vasconcelos 2025). Muriquis play a critical ecological role as folivorous–frugivorous primates. Although their diet includes both leaves and fruits, they rely on a diverse range of plant resources, and dietary composition varies seasonally according to resource availability (Milton 1984; Vasquez et al. 2025). Consequently, changes in plant community composition or fruit phenology may directly affect the resource availability for these primates. Climate projections indicate that habitat suitability may decline for 48–58% of Atlantic Forest angiosperms by 2050, while nearly 80% of endemic tree species are already threatened with extinction (Leão et al. 2021; de Lima et al. 2024). Such reductions in plant diversity and distribution may further exacerbate the vulnerability of muriqui populations.

Our main objective was to evaluate how projected climate change may restructure the spatial coupling between muriqui species and their main food resources. While previous studies have separately modeled future suitability for Atlantic Forest primates and plants, none has explicitly assessed how climate-driven shifts may alter the spatial congruence between primates and their main food plants. To address this gap, we (i) quantified changes in climatically suitable areas for muriquis. and their key food plants, (ii) evaluated shifts in the spatial overlap between primates and their resources, and (iii) estimated changes in the number of food species available per pixel under future emission scenarios. Additionally, we characterized the projected changes in key bioclimatic variables across their geographic ranges to interpret patterns of suitability loss, gain, or displacement. We hypothesized that climate change will not only reduce suitable areas but also decouple primates from their food resources, decreasing local food availability and intensifying extinction risk (Leão et al. 2021; Lima et al. 2022; Portillo 2021; de Oliveira et al. 2023a, b; Murakami et al. 2023; Pompeu and Portella 2023; Pinto et al. 2023; Vasconcelos 2025).

Study area and target species

We delimited the study area based on the polygons provided by the IUCN Red List (IUCN 2022b) for B. arachnoides and B. hypoxanthus (Fig. S1). Both species are endemic to the Atlantic Forest biome of southeastern Brazil, a global biodiversity hotspot characterized by high levels of endemism and severe habitat loss (Myers et al. 2000). The current distribution of B. arachnoides encompasses the eastern São Paulo state, southern Rio de Janeiro, and small portions of northeastern Paraná and southern Minas Gerais states, primarily associated with the Serra do Mar mountain range. Brachyteles hypoxanthus occurs further north, in the eastern Minas Gerais, Espírito Santo, and northern Rio de Janeiro states, with a small disjunct population near the Minas Gerais–Bahia border, largely associated with the Serra da Mantiqueira and adjacent regions. The region is characterized by a predominantly humid subtropical to tropical climate, with marked precipitation seasonality and altitudinal gradients that influence temperature and rainfall patterns. Spatial analyses were conducted separately for each species within their respective IUCN-defined ranges.

Main food resources

We listed records of plant species used by muriquis as food resource (including fruits, leaves, flowers, and other plant items). These records were compiled and taxonomically harmonized by Vasquez et al. (2025), and we further complemented them with grey literature (theses, dissertations, and conference meetings) by screening the curriculum (Brazilian Lattes Platform, https://lattes.cnpq.br/) of the authors with the highest number of peer-reviewed publications containing data on food resources for muriquis. These records were derived from studies conducted across eight sites for B. arachnoides and two sites for B. hypoxanthus. Plant taxonomy was first harmonized using the flora package in R (Carvalho 2022), based on the Brazilian Flora Checklist (Flora e Funga do Brasil 2023), which updates species names and corrects spelling errors (90% similarity threshold). Second, records not resolved with the first procedure were individually checked and standardized using the Plants of the World Online database (POWO 2026).

We found 17 articles, one dissertation, and two conference meeting studies (325 plant records) for B. arachnoides and nine articles and three dissertations (113 records) for B. hypoxanthus (Table S1). To ensure comparability among studies with different sampling efforts, we defined main food resources as: (1) plant species with the highest frequency of records in the literature; (2) plant species with a proportion of consumption or time spent on consumption > 1%, used to exclude rare or occasional items and retain consistently used resources (based on proportional values relative to study effort, rather than absolute values); and (3) plant species qualitatively reported by the authors as an important source of food for these primates during any period of the year. In total, 46 plant species were considered (B. arachnoides = 26; B. hypoxanthus = 20) (Table S2).

Occurrence records

We used muriquis occurrence records compiled by Pinto et al. (2023) and updated them to June 2023 through a Web of Science search (https://www.webofscience.com) using Brachyteles as keyword. Additional records were obtained from Neotropical Primates volumes (2019–2023), as this specialized primatological journal frequently publishes occurrence records but is not indexed by the Web of Science.

We accessed occurrence records for 46 muriquis main food resources (plant species) available on SpeciesLink (Canhos et al. 2022) in February 2024. For Vochysia saldanhana, we complemented the SpeciesLink dataset (54 records) with additional occurrence data from the Global Biodiversity Information Facility (GBIF.org 2024). Only taxonomically accepted records by SpeciesLink, dated post-1980 (to ensure data accuracy), and in South America were used for modeling (see below).

Recent occurrence records of B. arachnoides from Paraná State (Hack et al. 2022) were included in model calibration. However, suitability projections were spatially restricted to the species’ geographic range defined by the International Union for Conservation of Nature (IUCN 2022b). Currently, this range does not spatially encompass the region of newly reported localities, which consequently lie outside the mapped suitability outputs.

Suitability models

We modeled climatic suitability for B. arachnoides and B. hypoxanthus using the current bioclimatic variables from WorldClim – Global Climate database (Fick and Hijmans 2017; version 2.1; 2.5 arc minutes resolution, ~ 5 km) and independent occurrence records (one record randomly selected per pixel). The 19 bioclimatic variables are derived from monthly temperature and precipitation data and represent annual trends (e.g., mean annual temperature, annual precipitation), seasonality (e.g., annual ranges in temperature and precipitation), and extremes (e.g., temperature of the coldest and warmest months, and precipitation of the wettest and driest quarters). Bioclimatic variables were spatially cropped, with the extent defined by the sum of the species’ geographic ranges (IUCN 2022b) with a buffer based on the species’ maximum potential dispersal. This followed Bowman et al. (2002), who demonstrated an isometric relationship between dispersal distance and home range size, providing a practical basis for estimating dispersal potential within a generation time. The maximum potential dispersal projected up to 2100 (mean year of the interval of future climate data, see below) was ~ 387 km for B. arachnoides and ~ 369 km for B. hypoxanthus (Pinto et al. 2023). Multicollinearity among bioclimatic variables was evaluated using the Variance Inflation Factor (VIF; Quinn and Keough 2002) from 10,000 random pixels in each species’ spatial extension, iteratively removing variables with the highest VIF until all values were < 3 (routine: https://github.com/oliveirab/R-codes/blob/master/vif_func.R) (Zuur et al. 2010). We built suitability models using the maximum entropy algorithm in MaxEnt 3.3.3k (Phillips et al. 2006; Hijmans et al. 2023) implemented in R software (R Core Team 2022). We calibrated 60 models using different combinations of feature classes (linear, quadratic, hinge, product, and threshold) and regularization multiplier (0.5–5.0). We evaluated model performance and selected the most suitable based on the lowest corrected Akaike Information Criterion (AICc), retaining only models with Area Under the ROC Curve (AUC) values higher than 0.7, given its widespread use in presence-only modeling and its ability to ensure comparability with previous studies (Muscarella et al. 2014; Burnham and Anderson 2004).

For plant species suitability modeling, we used the same bioclimate layers from WorldClim (see above) and included soil layers (SoilGrids™: version 2.0, 5 km resolution) (Hengl et al. 2017). SoilGrids™ provides global maps of soil properties, including pH, soil organic carbon content, bulk density, coarse fragment content, particle size distribution (sand, silt, and clay), cation exchange capacity, total nitrogen, and soil organic carbon stocks, mapped at six standard depth intervals (Hengl et al. 2017). Soil data are important predictors of plant species distribution and can improve model accuracy (Zuquim et al. 2022; Velazco et al. 2017). We aligned soil layers to the projection, resolution, and extent of bioclimatic layers, using the raster::resample function (Hijmans 2023) in R software (R Core Team 2022).

The bioclimatic and soil layers were organized within a spatial extent defined for each plant species, defined by the sum of a minimum convex polygon generated using the species’ occurrence records with a 50 km buffer (the maximum potential dispersal projected until 2100). We chose this value because plant dispersal rates higher than 0.5 km/year are unrealistic (Warren et al. 2013; Lima et al. 2022). We used VIF to avoid multicollinearity among bioclimatic layers (as previously reported for primates). We performed a Principal Component Analysis (PCA) to reduce the dimensionality of soil layers (hereafter PCA soil layers), randomly selecting 10,000 pixels. Axes explaining more than 10% of the variance and cumulatively accounting for over 60% of the total variance (Hirzel and Le Lay 2008) were spatially projected and used in suitability models. We built suitability models using MaxEnt 3.3.3k and selected models using AIC and AUC (as previously reported for primates).

We spatially projected suitability models for each species (primates and resources) using current and future bioclimatic data from WorldClim, for 2041–2060 and 2081–2100 intervals (hereafter 2060 and 2100). We used bioclimatic variables derived from all 10 Atmosphere-Ocean General Circulation Models (AOGCMs) available during the data acquisition period (August 2023–March 2026: ACCESS-CM2, CMCC-ESM2, EC-Earth3-Veg, GISS-E2-1-G, INM-CM5-0, IPSL-CM6A-LR, MIROC6, MPI-ESM1-2-HR, MRI-ESM2-0, and UKESM1-0-LL) under four greenhouse gas emission scenarios (Shared Socio-economic Pathways — SSP; from lowest to highest emissions: SSP1-2.6, SSP2-4.5, SSP3-7.0, and SSP5-8.5) (CarbonBrief 2024). After careful data exploration (Booth 2022), we excluded BCC-CSM2-MR bioclimatic layers due to discrepant precipitation data compared with the other AOGCMs. For plant species, PCA soil layers were combined with future bioclimatic layers to spatially project suitability models. For each species (primates and resources), after spatially projecting models using future bioclimatic data, we built an average model for each SSP, considering all AOCGMs.

All modeling and projection procedures resulted in nine continuous suitability surfaces (ranging from zero to one) for each primate species and its food resources. This included one current and eight future suitability surfaces, a combination of each greenhouse gas emission scenario (SSPs) with two future intervals (2060 and 2100). We cropped suitability surfaces for each primate species’ geographic range (IUCN 2022b). We transformed continuous suitability surfaces into binarized surfaces (henceforth, distribution) using the 10th percentile of suitability values (minimum predicted suitability value that included 90% of the training locations) as a threshold for each species (Escalante et al. 2013).

Data analysis

To characterize the spatial patterns of climate change across the geographic ranges of muriquis (IUCN 2022b) and identify the bioclimatic variables that most influenced these variations, we conducted a PCA. We cropped the 19 bioclimatic layers from WorldClim (current and future projections from nine AOGCMs) to the geographic range of the muriquis. For each AOGCM and bioclimatic variable, we calculated the difference between future and current values and averaged these differences across the AOGCMs. Analyses were performed for four greenhouse gas emission scenarios (SSP1-2.6, SSP2-4.5, SSP3-7.0, and SSP5-8.5) and two future intervals (2060 and 2100). For PCA, we randomly selected 1,000 points from the average difference layers.

For each muriqui species and its respective set of main food plants, we calculated the projected distribution area (hectares) under five climate scenarios (one current and four future scenarios). We then assessed, for each primate species, whether the projected plant distribution area (response variable) differed between climate scenarios (independent variable) using the Friedman test (plant species as blocks). For each climate scenario, we measured the area of overlap between the projected distribution of primates (consumers) and that of each plant species (food resources) (Fig. 1). We assessed whether the overlap area (response variable) varied among current and future climate scenarios (independent variable) using the Friedman test (plant species as blocks). For each climate scenario, we calculated the number of resource species per pixel by summing all plant distributions in each pixel (Fig. 1). We assessed whether the number of plant species used as resources per pixel (response variable) varied among current and future climate scenarios (independent variable) using the Friedman test (1,000 randomly selected pixels used as blocks). For all Friedman tests, effect sizes were estimated using Kendall’s W, and when significant, pairwise post-hoc comparisons were performed using the Nemenyi test. We identified regions of resource loss and gain by calculating the difference (∆)between the future (each scenario) and current number of plant species per pixel. For each future scenario, we assessed whether these differences (∆) statistically differed from zero using the Wilcoxon signed-rank test (1,000 randomly selected pixels), a nonparametric equivalent of the one-sample t-test. This test can indicate negative (lower number of plant species in the future than in current scenarios) or positive (higher number of plant species in the future than in current scenarios) changes in the future.

Computational Workflow and Reproducibility

All analyses were conducted in R software version 4.2.1 (R Core Team, 2022). Species distribution models were implemented using the packages raster (version 3.6–26; Hijmans 2023), dismo (version 1.3–14; Hijmans et al. 2023), and ENMeval (Kass et al. 2021). Spatial vector operations were performed using sf (Pebesma, 2018). The areas were calculated using landscapemetrics with the function lsm_c_ca() (Hesselbarth et al. 2019). Bioclimatic and soil raster layers were processed using functions from the raster package, including crop(), mask(), stack(), extract(), reclassify(), resample(), predict(), and sampleRandom(). Model calibration, tuning, and evaluation were conducted using ENMeval 2.0 through the ENMevaluate() framework, with model selection based on the performance metrics and information criteria. Maxent models were fitted using the maxent() function implemented in the dismo package. Maps were created using the Free and Open Source QGIS (QGIS Development Team, 2022).

Bioclimatic conditions shifted between current and projected climate scenarios (2060 and 2100), particularly in extreme scenarios (Table S3, Axes 1–2). Future shifts projected under PCA 1 are related to precipitation (mainly annual precipitation, represented by the sum of the monthly averages) and temperature variability (mean diurnal range and temperature seasonality), while those under PCA 2 are related to temperature (minimum temperature of coldest month and mean temperature of coldest quarter) (Table S3). Precipitation was projected to increase in the range of B. arachnoides, and to decrease in the range of B. hypoxanthus (Fig. S2). Temperature variability will rise in the northern part of the range of B. hypoxanthus, and overall temperatures are projected to increase across the ranges of both species, especially in the western regions (Fig. S2).

Across taxa (both muriqui species and plant species), projected patterns of climatic suitability change were consistent across all emission scenarios and time intervals analyzed (2060 and 2100), with progressively larger reductions under higher emission pathways and toward 2100. Given this consistent pattern, all subsequent results are presented following the same structure, focusing on current and two future scenarios (the intermediate SSP2-4.5 and high-emission SSP5-8.5) for 2060 to enhance clarity and avoid redundancy. The complete results for all scenarios and projected time periods are provided in the Supplementary Material.

The selected suitability model showed excellent predictive performance (SI 2). For B. arachnoides, Isothermality (percentual contribution: 55.4), Temperature Annual Range (27.1), and Precipitation of the Wettest Month (13.9) mostly influenced the selected model. High suitability values (> 0.75) were more frequent under current conditions (Fig. 2). Spatially, the southern and central regions of B. arachnoides’ range had the highest suitability values in current and 2060-future scenarios (Fig. 2A-C). Future scenarios showed a decrease in the frequency of highly suitable pixels and an increase in intermediate values, especially under higher emissions, with a 4.5% reduction in the number of highly suitable (> 0.75) pixels under 2060-SSP2-4.5 and 27.4% under 2060-SSP5-8.5 (Fig. 2D–F). Projected distributions (suitability threshold > 0.18) for 2060-future were similar to or slightly larger than the current one (Fig. 2G-I), with a maximum 15% increase in the highest greenhouse gas emission scenario (SSP5-8.5), and no spatial shifts predicted (Fig. 2G-I).

For B. hypoxanthus, the Mean Temperature of the Wettest Quarter (percentual contribution: 45.1), Precipitation of the Wettest Month (31.3), Isothermality (12.4) and Precipitation of the Coldest Quarter (11.4) mostly influenced the selected model. High suitability values (> 0.75) were more frequent in the current scenario (Fig. 2), spatially concentrated in the central and eastern regions of the range (Fig. 2J). In 2060-future scenarios, high suitability regions nearly disappeared, with intermediate values in the east (Fig. 2J–L, M–O). Projected distribution (threshold > 0.15) declined by up to 67% under the highest greenhouse gas emission scenario (SSP5-8.5) (Fig. 2P-R), with the remaining suitable areas largely restricted to the eastern portion of the species’ distribution (Fig. 2P-R). Trends were consistent across other scenarios and the 2100-future interval, with larger changes under higher emissions by 2100 (Fig. S3).

Suitability models for all food resources had satisfactory AUC values (> 0.75; median = 0.902) (Table S4). Median suitability values of B. arachnoides (N = 26 food plant species) and B. hypoxanthus resources (N = 20 food plant species) were higher in current scenarios than in 2060-future ones for 16 (Fig. 3A) and 17 plant species, respectively (Fig. 3B). Median suitability values were lower under higher-emissions scenarios (Fig. 3). Trends were consistent across other scenarios and the 2010-future interval, with larger changes under higher emissions and by 2100 (Fig. S4). Differences described here refer to simple contrasts among model projections rather than formal hypothesis testing.

Projected distributions of B. arachnoides resources differed between current and 2060-future scenarios (Friedman test: χ² = 14.8, p < 0.01). Small spatial reductions were observed under future projections, with median decreases ranging from 0.3% (SSP2-4.5) to 1.4% (SSP5-8.5) (Table S5). Most resource distributions were projected to contract, while some plants, such as Copaifera trapezifolia and Cordia sellowiana were projected to remain stable (Table S5). Largest contractions were projected for Didymopanax angustissimus (28%), Aspidosperma polyneuron (25%), Eugenia involucrata (22%) and Pereskia aculeata (14%) in the SSP5-8.5 scenario (Table S5). Overlap between the distributions of B. arachnoides and its food resources differed among current and 2060-future scenarios (Friedman test: χ² = 13.4, p < 0.01), with higher values in SSP5-8.5 scenario (median = 5.3 million ha) than current ones (median = 4.9 million ha; Fig. 4B, Table S6).

Projected distributions of B. hypoxanthus resources differed among current and 2060-future scenarios (Friedman test: χ² = 22.5, p < 0.01). Median reductions ranged from 10.3% (SSP2-4.5) to 15.2% (SSP5-8.5). Largest contractions were projected for Didymopanax calvus (48%), Hirtella martiana (44%), Ocotea glaziovii (39%), Tovomita fructipendula (37%), Helicostylis tomentosa (36%), Adenocalymma marginatum (34% in SSP5-8.5; Table S5). Overlap between the distributions of B. hypoxanthus and its resources differed among current and 2060-future scenarios (Friedman test: χ² = 37.3, p < 0.01), with smaller values in SSP5-8.5 scenario (median = 800 thousand ha) than in current ones (median = 3.1 million ha; Fig. 4D, Table S6). Trends were similar across other scenarios and 2100, with larger reductions under higher emissions (Fig. S5, Table S6).

The number of plant species per pixel for B. arachnoides was higher in the current scenario than in 2060-future projections (Friedman test: χ² = 1359.0, p < 0.01), with greater reductions under higher-emissions scenarios (Fig. S6). Regions with the lowest number of plant species per pixel under current conditions (southern and northeastern portions of the range) maintained this pattern in future scenarios (Fig. 5). Difference in the number of food resources per pixel (future – current) was consistently negative in all future scenarios (p < 0.01; Fig. S6 for statistical tests), indicating resources loss, with largest reductions projected in southern and northeastern regions under high-emission scenarios (Fig. 5, S6).

For B. hypoxanthus, the number of plant species per pixel was higher under current scenario than in 2060-future projections (Friedman test: χ² = 1673.0, p < 0.01), with greater reductions under higher emissions (Fig. S6). Regions with the lowest number of plant species per pixel (northern and central portions of the range) remained low in future scenarios, while northeastern and southwestern areas showed no loss (Fig. 5). Differences in the number of food resources per pixel (future – current) were consistently negative (p < 0.01; Fig. S6 for statistical tests), indicating future resources loss, particularly in higher-emission scenarios (Fig. 5, Fig. S6). The northeastern and southwestern regions of the range had no projected loss of resource species. Trends were similar for other scenarios and 2100, with larger reductions under higher emissions (Figs. S6–S7).

Our results indicate that climate change will reduce climatic suitability for both muriqui species and their main food resources across their geographic ranges, with asymmetric impacts between species. While B. arachnoides showed relatively stable projected future distributions despite reductions in suitability, B. hypoxanthus was projected to experience substantial distribution contractions under future climate scenarios, particularly under high-emission pathways. In addition, the spatial overlap between muriquis and their main food resources is projected to decline, primarily due to the contraction of climatically suitable areas for primates rather than pronounced losses in suitability for plant species. Overall, our findings indicate that climate change may intensify conservation risks for B. hypoxanthus populations, whereas B. arachnoides is projected to undergo comparatively moderate distributional changes.

Previous studies have consistently reported reductions in climatically suitable areas for muriqui species. For B. hypoxanthus, projected declines range from 24 to 43% by 2060 and from 44 to 39% by mid-century and from 13 to 69% by 2100 (Portillo 2021; Pinto et al. 2023; Pompeu and Portella 2023; Vasconcelos 2025), although one study reported a slight increase (~ 5%) in future distribution (de Oliveira et al. 2023a, b). In comparison, our results indicate smaller mid-century reductions for B. arachnoides but more pronounced declines for B. hypoxanthus under high-emission scenarios (~ 66%). In addition to range shifts, the general reduction in quantitative suitability from highly suitable pixels and throughout the range is noteworthy. This is the first study to report this qualitative change in climatic suitability throughout the range of muriquis.

While all studies converge in projecting reductions in climatically suitable areas, the differences in magnitude likely reflect variations in the modeling approaches. Earlier studies (e.g., Portillo 2021; Pompeu and Portella 2023) used Representative Concentration Pathways, whereas our study employed updated Shared Socioeconomic Pathways, which may lead to different projections of future climatic conditions (CarbonBrief 2024). Additionally, differences in climate data sources (e.g., CHELSA vs. WorldClim) are known to influence model outputs (Bobrowski et al. 2021). The extent of model calibration is another important source of variation. For example, Pinto et al. (2023) projected higher reductions for B. arachnoides (44–53% by 2060 and 50–74% by 2100), likely associated with the use of broader calibration extents that included all Atlantic Forest primates combined with dispersal buffers. Broader calibration areas increase environmental heterogeneity and may enhance the contrast between presence and background conditions, influencing binarization thresholds and potentially leading to higher estimates of range contraction and commission errors (Escalante et al. 2013; Vasquez et al. 2021). Despite these methodological differences, all studies consistently indicate a declining suitability for both species. Spatially, area reductions for B. arachnoides are projected across both the northern and southern portions of its range, whereas for B. hypoxanthus, the area contraction is projected toward the central-eastern region of its current distribution.

The distributions of the main food resources of muriquis are also projected to decrease under climate change, with higher reductions for B. hypoxanthus resources. This indicates a contraction in the spatial availability of key food resources, particularly under higher emission scenarios. A small subset of plant species is projected to maintain or slightly expand their suitable areas under at least one mid-century scenario, including Coussapoa microcarpa, Esenbeckia leiocarpa, Mabea fistulifera, and Xylopia brasiliensis for B. arachnoides, and Virola bicuhyba and Spondias dulcis for B. hypoxanthus. These species are documented as recurrent food resources and contribute to fruit availability during specific seasonal windows, but their contribution is associated with specific seasonal periods; thus, their persistence does not imply continuous resource availability throughout the year. However, most expanding or stable species do not compensate for the substantial contractions projected for several other frequently consumed resources, particularly those with currently restricted distributions. Species predicted to expand are characterized by relatively broad climatic tolerance and wide geographic ranges, which are traits generally associated with greater resilience to climate change. Species with broader climatic niches and larger range sizes tend to be less negatively affected by shifting environmental conditions, whereas narrow-range species are often more vulnerable (Zwiener et al. 2017; Vincent et al. 2020; Leão et al. 2021). As a result, food resources are projected to become more spatially restricted and fragmented, particularly for B. hypoxanthus, indicating that gains in the distributions of some species do not offset the loss of others.

Climate change is projected to strongly affect the spatial overlap between muriquis and their main food resources, with more pronounced impacts for B. hypoxanthus than for B. arachnoides. This difference is consistent with the greater loss of climatic suitability projected for both B. hypoxanthus and its associated food resources in the northern portion of the Atlantic Forest compared to the southern regions where B. arachnoides occurs. Projected southward shifts in plant distributions (Colombo and Joly 2010) further support this pattern, as they may reduce the spatial overlap between B. hypoxanthus and its current food resources while having less pronounced effects for B. arachnoides. These shifts suggest that climate change may alter not only species distributions but also the spatial alignment between consumers and their resources.

Our results also indicated a reduction in the number of currently documented food plant species per pixel, particularly for B. hypoxanthus, suggesting decreased resource availability. Given that muriquis rely on a diversity of fruit resources to track seasonal availability (Milton 1984), reductions in the number of available species may affect the temporal continuity of food supply. These estimates are based on currently known food resources and do not account for their nutritional composition, how muriquis balance nutrient intake, or the potential incorporation of new species into the diet under future conditions. As plant species shift their distributions, functionally similar resources may become available in newly suitable areas. In this context, partial compensation could occur if newly available plant species replace currently consumed resources with similar ecological characteristics, such as fruit accessibility, phenology, and nutritional value. Nevertheless, the extent to which such replacement may occur remains uncertain and depends on dispersal processes, establishment success, and maintenance of phenological complementarity. In parallel, projected changes in plant communities, particularly the expansion of widespread species and the decline of range-restricted ones (Zwiener et al. 2017), may reduce the diversity of available resources.

Seasonal climate variations, habitat disturbance, and reduced resource availability can alter primate behavior (Dib et al. 1997; Mourthé et al. 2007). The direct effects of high temperatures can potentially result in physiological responses such as heatstroke and heat exhaustion (Pozo-Montuy et al. 2024). Although Atelidae species are highly arboreal and adapted to suspension locomotion (Campbell et al. 2005), B. hypoxanthus individuals have been observed foraging, resting, and engaging socially near the ground (Dib et al. 1997; Mourthé et al. 2007), increasing vulnerability to hunting and predation, as well as enhancing pathogen exposure (Chapman et al. 2005; Mourthé et al. 2007).

Previous studies have modeled climatic suitability to estimate the future distributions of dietary plant resources of two Leontopithecus species (Raghunathan et al. 2015) and to identify suitable habitats for Sapajus libidinosus reintroductions (Donnini et al. 2024). To our knowledge, this is the first study to integrate key food resources and climatic suitability for both resources and consumers. Projected reductions in climatically suitable areas for muriquis and their resources, particularly for B. hypoxanthus, are an additional threat to these primates due to climate change. Both species are classified as Critically Endangered, primarily due to deforestation, fragmentation, and hunting (de Melo et al. 2021; Talebi et al. 2021), with forests supporting viable populations limited to approximately 8% (B. hypoxanthus) and 13% (B. arachnoides) of their historical ranges (Ingberman et al. 2016). However, our models were based on climatic suitability and did not explicitly integrate landscape configuration or habitat availability, factors that may impose additional constraints on future persistence (see Rezende et al. 2020). Given the already severe degree of habitat fragmentation across the Atlantic Forest, the combined effects of further climatic contraction and landscape constraints may exacerbate population isolation and reduce long-term viability.

Implications for Conservation

Our projections indicate a contraction and spatial reorganization of climatically suitable areas for both B. hypoxanthus and B. arachnoides, with stronger declines for B. hypoxanthus under high-emission scenarios. This shift raises concerns about the alignment between suitable future areas and the current protected area network, which was largely established based on historical distributions (Hannah et al. 2007). Although most known muriqui populations occur within legally protected areas (Strier et al. 2017), these reserves may not fully encompass zones projected to remain suitable in the coming decades. Therefore, conservation planning should incorporate climate projections to identify and prioritize areas of long-term climatic stability, particularly in montane regions that may function as climatic refugia and areas where muriquis mostly occur (Dobrowski 2011; Morelli et al. 2016). However, because such areas are frequently embedded in highly fragmented landscapes within the Atlantic Forest, ensuring functional connectivity is critical. Muriquis are dependent on large, continuous, and structurally complex forests (de Melo et al. 2021; Talebi et al. 2021), and their ability to disperse to climatically suitable areas will depend on maintaining enriching forest fragments and restoring landscape connectivity. Ecological corridors can ensure that this dispersal occurs; however, they should be planned considering the projected climatic stability, existing forest cover, elevational gradients, and spatial distribution of key food plant species. Previously proposed corridors, such as RPPN Mata do Sossego–RPPN Feliciano Miguel Abdala (Caratinga) and PE Desengano–PN Serra dos Órgãos (Strier et al. 2017), represent strategic starting points for connecting current populations; however, their design should be reassessed under future climate scenarios to ensure they effectively link areas projected to remain suitable.

Projected reductions in suitable areas for key food plant species indicate potential declines in the spatial overlap between muriquis and important dietary resources. Although dietary shifts, including increased folivory or incorporation of alternative fruit species, are possible due to the behavioral plasticity of muriquis, the more pronounced contraction in climatically suitable areas for the primates themselves represents the primary conservation concern. Moreover, dietary data are geographically concentrated. For B. hypoxanthus, long-term studies are restricted to RPPN Feliciano Miguel Abdala (Caratinga) and RPPN Mata do Sossego, while information for B. arachnoides is unevenly distributed across its range (Vasquez et al. 2025). This limited and spatially uneven sampling may influence the representation of plant species in the dataset, particularly with respect to local variations in resource use. Nevertheless, the compiled dataset is based on the best available evidence and includes recurrently consumed species that capture the core dietary patterns of both species. Our projections provide a consistent and ecologically meaningful representation of resource availability at the species level, while acknowledging that finer-scale regional variation may not be fully represented. Therefore, expanding dietary and phenological research across the full geographic range of both species is essential to identify region-specific key plant species, guide habitat enrichment, and refine assessments of trophic vulnerability under climate change.

Overall, conservation strategies for muriquis must integrate climate projections, trophic dynamics, and landscape configuration. Priority actions include (1) evaluating future climatic suitability for all known populations, (2) expanding dietary and phenological studies across the species’ entire distributions, (3) reassessing and implementing ecological corridors that connect populations in climatically declining regions to more stable areas, and (4) enriching occupied forest fragments with key food plant species to enhance long-term resource stability. Without climate-informed, connectivity-oriented planning, current conservation frameworks may be insufficient under accelerating climate change.

Authors

1. Vagner Lacerda Vasquez
2. Míriam Plaza Pinto

Cite this article

Vasquez, V.L., Pinto, M.P. Matching the largest Neotropical primates with their food resources: Rapid suitability declines in light of climate change. Biodivers Conserv 35, 192 (2026). https://doi.org/10.1007/s10531-026-03396-8

Download citation

• Received: 10 December 2025

• Revised: 12 June 2026

• Accepted: 16 June 2026

• Published: 25 June 2026

• Version of record: 25 June 2026

• DOI: https://doi.org/10.1007/s10531-026-03396-8

Keywords