Environmental heterogeneity as a determinant of bee diversity patterns in the Atlantic Forest

Anderson Igomar Antonio; Admir Cesar de Oliveira Junior; Fabricio Villalobos; Santiago José Elías Velazco

doi:10.21425/fob.18.142410

Research Article

Environmental heterogeneity as a determinant of bee diversity patterns in the Atlantic Forest

Anderson Igomar Antonio^‡, Admir Cesar de Oliveira Junior^‡, Fabricio Villalobos^§, Santiago José Elías Velazco^|¶

‡ Universidade Federal da Integração Latino-Americana, Foz do Iguaçu, Brazil

§ Red de Biología Evolutiva, Instituto de Ecología, Xalapa, Mexico

| San Diego State University, San Diego, United States of America

¶ Universidad Nacional de Misiones, Puerto Iguazú, Argentina

Corresponding author: Santiago José Elías Velazco ( svelazco@sdsu.edu )

Academic editor: Marcus Vinicius Cianciaruso

This is an open access article distributed under the terms of the Creative Commons Attribution License (CC BY 4.0), which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.

Citation: Antonio AI, Oliveira Junior ACde, Villalobos F, Velazco SJE (2025) Environmental heterogeneity as a determinant of bee diversity patterns in the Atlantic Forest. Frontiers of Biogeography 18: e142410. https://doi.org/10.21425/fob.18.142410

Abstract

The Atlantic Forest encompasses a wide range of environmental and geographical gradients with high endemism and species diversity among several taxonomic groups, including bees. Environmental heterogeneity is a determining factor for species diversity, as environments with greater heterogeneity tend to offer a greater variety of conditions, thus supporting higher species richness. However, bee richness patterns and their relationship with environmental heterogeneity in the Atlantic Forest remain underexplored. In this study, we aimed to describe the bee diversity patterns and investigate how different components of environmental heterogeneity—specifically temperature seasonality, topographic and geomorphic heterogeneity, and stream density—influence species richness, both for the entire biome and within each ecoregion. To do so, we modeled and estimated the distribution of 466 bee species. Relationships between bee species richness and environmental heterogeneity variables were analyzed using Generalized Linear Models, variable importance, and partial dependence curves. We found that the highest richness was in the southwestern regions of the Atlantic Forest, particularly in the Serra do Mar Coastal Forests and Araucaria Moist Forests. The most important variables positively related to species richness were temperature seasonality, followed by topographic and geomorphic heterogeneity, whereas stream density showed the lowest importance. At the ecoregion level, temperature seasonality was the most important variable for 9 of the 11 ecoregions, followed by topographic and geomorphic heterogeneity. In ecoregions with the highest bee richness, environmental heterogeneity showed a low explanatory power. Notably, the relationships between the environmental heterogeneity variables and species richness varied across ecoregions. Our findings highlight the significant role of environmental factors in shaping bee species richness in the Atlantic Forest at multiple scales. Furthermore, the distinct relationship observed between environmental heterogeneity and species richness across ecoregions reinforces the necessity of multi-scale diversity studies to elucidate the unique characteristics of each ecoregion.

Highlights

Bee species richness in the Atlantic Forest is greatest in the south-west and southern regions.
Seasonal temperature was the most important variable for predicting species richness, showing a positive effect in nine of the 11 ecoregions.
Topographic and geomorphic heterogeneity contributed significantly to species richness in certain ecoregions.
Relationship between species richness and environmental heterogeneity variables varied among ecoregions.
Ecoregions with higher richness were poorly explained by environmental heterogeneity, suggesting other influencing factors.
Multiscale analysis reveals that different regions may have unique species richness drivers, crucial to understanding species diversity in this biodiversity hotspot.

Keywords

Apoidea, Atlantic Forest, geomorphic heterogeneity, nonstationarity, richness pattern, seasonality, species distribution modeling, topographic heterogeneity

Introduction

The Atlantic Forest biome extends across eastern Brazil, northeastern Argentina, and eastern Paraguay. This region exhibits a wide range of latitudinal, altitudinal, and environmental gradients and comprises multiple ecoregions with distinctive environmental attributes (Olson et al. 2001; Batista et al. 2021; Carvalho et al. 2021). The Atlantic Forest is considered a biodiversity hotspot and is home to 2,200 vertebrate species (i.e., > 5% of the world’s species; Myers et al. 2000), of which 334 are mammals (Souza et al. 2019). In addition, 35% of 20,000 native plant species (Faoro et al. 2015), 16% of 688 bird species, 31% of 200 reptile species, and 60% of 280 amphibian species are endemic to the Atlantic Forest (Mittermeier et al. 2005). This biome represents a region of high biological importance for arthropods, with 724 areas of endemism identified (Hoffmeister and Ferrari 2016). Recent studies based on DNA Barcoding have revealed a great diversity of arthropod species, many of which have not yet been cataloged (Bukowski et al. 2022). The region exhibits remarkably high biodiversity of eusocial insects, including bees, wasps, and termites (Feitosa et al. 2021), with high bee endemism, particularly in the genus Euglossa (Garraffoni et al. 2017).

The Anthophila clade comprises all bee species and represents a highly diverse group within Apoidea (Michener 2007), with > 20,000 species described globally, whose diversity is reflected in the wide variation in behavior, ecology, and distribution (Michener 2007). Bees provide important ecosystem services, the most important of which is pollination (Michener 2007; Rogers et al. 2014). This ecosystem service is essential for the sexual reproduction of plant species (Ollerton et al. 2011) and has a direct effect on the genetic diversity of plant seeds (Ramos and Schiestl 2019). In addition, crop pollination can be maintained exclusively by native bees, especially if natural habitats near plantations increase (Kremen et al. 2004; Ricketts et al. 2008).

Studying diversity patterns is fundamental for understanding the dynamics and structure of ecological communities, as well as for helping manage and conserve biodiversity (Rubene et al. 2015). These patterns are influenced by both evolutionary and ecological factors, biotic and abiotic, at different spatial scales (Ricklefs 2004). Bee diversity patterns have been studied because of their ecological importance and pollination services (Klein et al. 2006). Previous research has shown that bee species diversity is influenced by a variety of biotic and abiotic factors, including availability of floral resources, habitat, interactions with other species, and climatic variations (Potts et al. 2003; Classen et al. 2015; Escobedo-Kenefic et al. 2020). Recently, bee distribution has been mapped globally, showing a bimodal latitudinal gradient with a higher concentration of species at mid-latitudes than at low latitudes (Orr et al. 2021). Although there have been several studies on animal and plant diversity in the Atlantic Forest (Jenkins et al. 2015; Souza et al. 2019), there is still a gap in the diversity patterns of bees at broad scales and the environmental factors related to those patterns.

Environmental heterogeneity is defined as the variation in biotic (e.g., vegetation structure) and abiotic (e.g., topography) environmental characteristics within a given area (Maliniemi et al. 2024). Environmental heterogeneity plays a key role in shaping biodiversity patterns by promoting niche differentiation and reducing interspecific competition (Whalen et al. 2016; Xu et al. 2016; Tukiainen et al. 2017). Variations in land cover, vegetation, climate, soil, and topography strongly correlate with species richness across taxa and spatial scales (Dufour et al. 2006; Stein et al. 2014; Wan et al. 2023). Highly heterogeneous environments support greater species diversity and offer varied conditions and resources for different niches and refuges during environmental change (Stein and Kreft 2015). Furthermore, from an evolutionary perspective, heterogeneous environments can induce allopatric speciation, restrict gene flow, and foster local adaptation (Dool et al. 2022). Some studies in the Atlantic Forest have explored the relationship between different aspects of species diversity and environmental heterogeneity (Stevens 2013; Weber and Cáceres 2018; Delciellos et al. 2022). However, these studies primarily focused on vertebrates and plants using topographic heterogeneity as the main variable. As such, the influence of environmental heterogeneity on bee diversity remains to be evaluated, especially in particularly diverse regions such as the Atlantic Forest and considering multiple environmental variables.

Bees are sensitive to environmental changes and have specific requirements for foraging, gathering resources, and maintaining their nests (Klein et al. 2017; Burdine and McCluney 2019). Therefore, environmental heterogeneity is expected to have a positive relationship with bee diversity because of the variety of environments and resources, thus supporting greater species richness. In macroecological studies, environmental heterogeneity can be measured using multiple metrics (Stein and Kreft 2015). For example, heterogeneity in soil chemical properties (Xu et al. 2016), geomorphic heterogeneity, or water richness (Tukiainen et al. 2017). Due to the longitudinal, latitudinal, and climatic amplitudes of the Atlantic Forest, different metrics of environmental heterogeneity are expected to affect the patterns of bee species richness. Temperature seasonality indicates variations in average monthly temperatures yearly (O’Donnell and Ignizio 2012), which can affect plant phenology, resource availability timing, and bee thermal limits. Topographic heterogeneity refers to elevation variability; greater variation indicates greater topographic heterogeneity (Amatulli et al. 2020). Additionally, geomorphic heterogeneity refers to the variation of topographic features measured by classes of geomorphological forms (Amatulli et al. 2020). High topographic and geomorphic heterogeneity creates diverse microclimates and environmental conditions. Areas with greater topographic heterogeneity are expected to have higher species richness, as such heterogeneity fosters different conditions for bees with different ecological requirements. Stream density represents the number of potential watercourses in an area and is associated with humidity and riparian habitats (Keeton et al. 2007). This variable is important for certain bee species because of the water needed for their survival and nest temperature control (Kühnholz and Seeley 1997). Thus, temperature seasonality, topographic and geomorphic heterogeneity, and stream density are expected to have a positive relationship with bee species richness.

The relationship between a response variable (e.g., species richness) and predictor variables (e.g., environmental heterogeneity) can vary significantly at different temporal or spatial scales, a phenomenon known as nonstationarity (Fortin and Dale 2005). Nonstationarity implies that ecological patterns and processes, and their relationships with different variables, are not constant over time or space (Rollinson et al. 2021). Because of the broad geographical and environmental extent of the Atlantic Forest, it is expected that there will be nonstationarity between bee richness and different components of environmental heterogeneity. In other words, these relationships could change between biome and different ecoregion scales, making it pertinent to explore these relationships at different levels separately (i.e., at the global level of the biome and within each ecoregion). In this study, we aimed to describe the patterns of native bee diversity in the Atlantic Forest and investigate how the different components of environmental heterogeneity influence bee richness in the Atlantic Forest on a global scale and in each ecoregion.

Material and methods

Study area

The Atlantic Forest extends along the entire eastern coast of Brazil, from Rio Grande do Norte to Rio Grande do Sul, and through the coastal and continental areas to southern Brazil, eastern Paraguay, and southwestern Argentina (Ribeiro et al. 2009; Muylaert et al. 2018). The Atlantic Forest is separated from the Amazon Rainforest by a corridor of seasonally dry forests and savannas such as semiarid Caatinga (northeastern Brazil), Cerrado (central Brazil), and Chaco (Paraguay, Argentina, and Bolivia) (Prado and Gibbs 1993). It contains a unique set of ecosystems that originally comprised > 1.5 million km² (Morellato and Haddad 2000). The biome has a wide latitude range (from 5°8'N to 33°8'S), an altitude between 0–2,200 m a.s.l., rainfall between 800–4,000 mm/year, and an average annual temperature ranging from 5–25 °C (Stehmann et al. 2009).

We used the integrative boundary of the Atlantic Forest (Muylaert et al. 2018) because it covers a more inclusive area, considering the transition regions with other biomes such as Cerrado and Caatinga, as well as representing portions of the Atlantic Forest for other countries, such as Argentina and Paraguay. We clipped the boundaries of the world’s terrestrial ecoregions from the polygon as the integrating boundary of the Atlantic Forest (https://github.com/LEEClab/ATLANTIC-limits) (Dinerstein et al. 2017, https://ecoregions.appspot.com). Initially, 19 ecoregions were listed within the Atlantic Forest boundaries; however, many of them were represented marginally or with small extensions, which could make it difficult to analyze them separately (Suppl. material 1: table S1, fig. S1). Polygons of marginal ecoregions (i.e., < 3,000 km² equivalent to 120 cells) were joined to the largest and closest ecoregions. A total of 11 ecoregions were analyzed: Araucaria Moist Forests, Bahia Coastal Forests, Bahia Inland Forests, Brazilian Atlantic Dry Forests, Caatinga, Campos Rupestres Montane Savanna, Cerrado, Pernambuco Coastal Forests, Serra do Mar Coastal Forests, Upper Paraná Atlantic Forests, and Uruguayan Savanna.

Species list and occurrences

To construct the species list for the Atlantic Forest, we followed two stages. First, a list of species occurring in the Atlantic Forest was generated from occurrences sourced by Dorey et al. (2023). This occurrence dataset provides bees occurrences worldwide compiled, integrated, and cleaned from Atlas of Living Australia (ALA, https://www.ala.org.au), Global Biodiversity Information Facility (GBIF, https://www.gbif.org), Symbiota Collections of Arthropods Network (SCAN, https://scan-bugs.org/portal/), Integrated Digitized Biocollections (iDigBio, https://www.idigbio.org), and the United States Geological Survey (USGS, https://www.usgs.gov) (Dorey et al. 2023). We integrated the Dorey et al. (2023) database with SpeciesLink (https://specieslink.net), which is a biodiversity information portal from Brazil.

The species list included all bee species with at least one occurrence in the Atlantic Forest. We then removed species without distribution in the Atlantic Forest. We used the Taxonomic Catalog of Brazilian Fauna (https://fauna.jbrj.gov.br/) and the Taxonomic Catalog of Brazilian Fauna and Moure Catalog (https://moure.cria.org.br/) as species origin sources. The scientific names were revised according to the Taxonomic Catalog of Brazilian Fauna and Moure Catalog. The raw occurrence data listed 770 species names, but after scientific name correction and selection of those species native to the study area, 564 native bee species to the Atlantic Forest were listed. Species with only one occurrence (n = 74) were eliminated, leaving 490 species for distribution modeling.

We performed a spatial cleaning that consisted of removing occurrences with invalid coordinates, duplicated coordinates, and georeferenced in centroids of municipalities, provinces/states, and countries. For data integration and cleaning, we used the R packages bdc (Ribeiro et al. 2022) and CoordinateCleaner (Zizka et al. 2019). For each species, we checked whether the occurrences were within the species’ natural distribution using the Moure Catalog as a source of species distribution. Occurrences georeferenced outside of the informed distribution were removed manually in a GIS environment using the program QGIS v. 3.38.0 (https://www.qgis.org/). Raw occurrence database contained 240,062 occurrences, which were reduced to 16,542 after cleaning.

Occurrences compiled from large databases generally have spatial bias, with a higher density of points close to human infrastructure (Stolar and Nielsen 2015). This bias can negatively affect the predictions of species distribution models (Baker et al. 2022). To correct sampling bias, we used environmental filtering (Varela et al. 2014). As this method is sensitive to the number of classes used to partition the environmental space (i.e., binds), for each species we tested 4, 6, 8, 10, and 12 classes and selected the filter with the lowest spatial autocorrelation (as measured by Moran’s I) and the largest number of retained occurrence records (Velazco et al. 2021) 11,404 occurrences were retained after this process.

Species distribution models and species richness maps

Environmental variables

Variation in temperature, humidity, and precipitation patterns can directly influence species behavior and resource availability across an environment, thus shaping the geographic distribution of species (Cortopassi-Laurino et al. 2007). Species distribution models were built using bioclimatic variables from Chelsa v2.1 (https://chelsa-climate.org) (Karger et al. 2017) at 1 km. Of the 19 bioclimatic variables, we selected 11 that were ecologically related to bee distribution (Suppl. material 1: table S2). We also used the 1 km resolution elevation from SRTM (https://srtm.csi.cgiar.org). The extent of the variables ranged from the northern border of the United States to the southern extremes of Argentina and Chile. All variables were resampled to a resolution of 5 km. A Pearson’s correlation matrix for the full range of environmental variables revealed significant collinearity among many of them (Suppl. material 1: fig. S2). To address multicollinearity and reduce the number of predictor variables used for modeling, we applied Principal Component Analysis (PCA) using raster cells from the training area of each species (see below). The PCAs were calculated based on a correlation matrix, and eigenvectors were used to predict principal components used as predictor variables in species distribution models. We selected a total of principal components that explained 95% of the total variance in the original environmental variables (De Marco and Nóbrega 2018). Thus, four, five, and six principal components were used for 282, 206, and two species, respectively.

Because the amount of occurrence data affects the performance of Species Distribution Models (SDM), and there are different techniques for dealing with lack of data, we defined three modeling protocols: one for species with ≥ 15 occurrences (n = 183 species), another for species with between 5 and 14 occurrences (n = 156), and another for those with between 2 and 4 occurrences (n = 151). Species with only one occurrence were not included in our analysis. For the first and second protocols, the SDM training area was defined as delimited by ecoregions where a species had at least one presence. We used Dinerstein et al. (2017; https://ecoregions.appspot.com) as the ecoregions polygon source for the 11 ecoregions. SDM were constructed using flexsdm R package (Velazco et al. 2022).

Modeling protocol for species with > 15 occurrences

No single algorithm can handle all modeling conditions, so we used the following seven algorithms: Artificial Neural Network (NET), Boosted Regression Trees (BRT), Generalized Additive Model (GAM), Generalized Linear Model (GLM), Maximum Entropy (MAXENT, hereafter MAX), Random Forest (RAF), Support Vector Machine (SVM), and Gaussian Process (GAU). The NET, BRT, MAX, RAF, and SVM algorithms have hyperparameters that can affect model performance and predicted suitability patterns (Fourcade 2021). Therefore, for these algorithms, we employed a hyperparameter optimization technique that selects the best combination of hyperparameters that maximizes the Sorensen performance metric (Suppl. material 1: table S3). The consensus was based on the median of the environmental suitability values because it is less sensitive to outliers (Rose et al. 2024).

Modeling protocol for species with 5–15 occurrences

For species with 5–15 occurrences, the Ensemble of Small Models (ESM) approach was used. This technique is suitable for building models for species with few occurrences and consists of creating bivariate models with a combination of all pairs of predictors and subsequent consensus between the bivariate models weighted by Somers’D metric (D = 2 × (AUC – 0.5)), where AUC is the area under the curve (Breiner et al. 2015). The ESMs were constructed using the same algorithms as in the previous protocol, however, the default algorithm hyperparameters were used (Velazco et al. 2022). As with SDM, the final models consisted of a consensus model based on the median.

Modeling protocol for species with 2–4 occurrences

For species with 2–4 occurrences, distributions were estimated based on environmental similarity using Gower’s distance for cells within a 50 km radius around the species occurrences (Carpenter et al. 1993; Andrella et al. 2023).

Model validation and modeling postprocessing

The SDM and ESM were validated using k-fold and repeated k-fold cross-validation techniques, respectively, with repetitions. Five partitions were used for SDM, and five partitions and five repetitions were used for ESM. We used the Inverse Mean Absolute Error –IMAE– (threshold-independent metric) and Sorensen (Leroy et al. 2018) as model performance metrics. Only those models with Sorensen ≥ 0.7 were used in the analyses. The 0.7 Sorensen value was employed as an acceptable threshold for model performance, as this value is widely recognized as a performance cutoff for other metrics such as Area Under the Curve (Swets 1988; Peterson et al. 2011).

When SDMs are projected over large geographical extents, models tend to predict high suitability areas outside the species’ current range, potentially affecting diversity patterns (Velazco et al. 2020). To address this, we constrained model prediction to a region defined by a minimum convex polygon based on a species presence plus a buffer of 100 km buffer around it (Mendes et al. 2020).

Species richness map

We created a species richness map by stacking species semi-binary models. The semi-binary models consist of assigning zero to environmental suitability values that are below the threshold and keeping all values above it continuous (Domisch et al. 2019). This approach to building the richness map was adopted because it reduces over-prediction compared to the sum of binary models (Guillera-Arroita et al. 2015). We used the threshold that maximizes the sum of sensitivity and specificity to produce the semi-binary model outputs. This threshold was used because it is little affected by the relationship between presences and pseudo-absences (Liu et al. 2016).

Environmental heterogeneity variables and analyses

Four environmental heterogeneity variables were explored: topographic heterogeneity, geomorphic heterogeneity, temperature seasonality, precipitation seasonality, and stream density. Topographic heterogeneity was based on a 30 m resolution digital terrain model from the R atlanticr package (https://github.com/mauriciovancine/atlanticr) and was calculated using the standard deviation of the altitudinal variation of the 30 m cells contained in the 5 km resolution cells (i.e., equal to the resolution of the distribution models). For geomorphic heterogeneity, we calculated the Shannon diversity of the different geomorphological features (plane, peak or summit, ridge, shoulder, spur, slope, hollow, slope, valley, and depression) of the 90 m resolution cells contained in 5 km resolution cells. We used Geomorpho90m as the geomorphological feature source (Amatulli et al. 2020). Stream density was measured by adding the number of cells identified as potential watercourses in 5 km resolution cells. For this, we used Strahler stream order at a 90 m resolution and considered watercourses for cells with stream order > 3. Hydrography90m was used as a source of the Strahler stream order (Amatulli et al. 2022). The seasonality of temperature and precipitation were sourced from Chelsa. Due to the high correlation between temperature and precipitation seasonality (Pearson’s correlation = -0.81), we selected the former variable for our analysis (Suppl. material 1: fig. S3).

The relationships between species richness and different environmental heterogeneity variables were analyzed for 12 regions: the Atlantic Forest as a whole and 11 ecoregions. For the extent of the Atlantic Forest, we also performed an analysis including ecoregions as a predictor variable together with environmental heterogeneity to evaluate the contribution of the ecoregion to explain the bee richness pattern. We used multiple regressions constructed using Generalized Linear Models (GLM) using the Poisson distribution family, which is suitable for discrete response variables (i.e., species richness). Assumptions of normality and homogeneity of the residuals were assessed visually. Moran’s I correlograms were used to assess the spatial autocorrelation of the residuals. To correct it, we used a spatial filter method based on eigenvectors (Tiefelsdorf and Griffith 2007). The SpatialFiltering function from the R package spatialreg (Pebesma and Bivand 2023) selects eigenvectors in a semi-parametric spatial filtering approach to remove spatial dependence from linear models (Pebesma and Bivand 2023). These spatial filters were used as predictor variables in the GLMs. Because the creation of environmental filters can be time-consuming, we systematically sampled ~2200 cells while maintaining the original resolution (i.e., 5 km) by constructing GLMs using a sample of the cells for each analysis area. To explore the variable importance and partial response curves estimated by the GLMs, we used the R packages spatialreg and vip to construct Moran’s I correlograms and spatial eigenvector filters, respectively (Greenwell and Boehmke 2020). emmeans package was used to calculate the variable importance and estimate the partial response curves (Lenth 2016). The method used to calculate variable importance is based on the absolute value of the t-statistic, which can vary from zero to infinity. For the response curves, we standardized the predictor variable scales to range between 0–1 and facilitate comparisons between different curves.

Temperature seasonality was used both to construct the SDMs and to analyze the relationship between richness and environmental heterogeneity, raising concerns about potential circularity—using the same variable in both modeling and explanatory analyses may artificially inflate the results. However, using one variable to model individual species distributions through SDMs is not the same as using this same variable to model the community-level pattern that emerges from stacking several individual modeled distributions (e.g., species richness; Ferrier and Guisan 2006). Indeed, the importance of such a variable in SDMs is more related to setting the geographical space potentially occupied by each species and thus its spatial limits, whereas the importance of this variable for species richness is more related to their spatial covariation (Ferrier and Guisan 2006; Guisan and Rahbek 2011). To assess potential circularity between predicted species richness and environmental heterogeneity, we examined correlations between individual species’ suitability and the environmental heterogeneity variables. For those species with Pearson correlation > |0.7|, we generated a richness map to explore if those species were highly concentrated in some region of the study area. All analyses were performed in R v.4.4.0 (R Core Team 2024).

Results

Ensemble SDM models performed well, with IMAE and Sorensen values of 0.73 (± 0.05) and 0.74 (± 0.08), respectively. Similarly, the ESMs presented a good performance, with average IMAE and Sorensen values between the algorithms of 0.63 (± 0.06) and 0.74 (± 0.11) (Suppl. material 1: fig. S4). We excluded 24 species because of poor model performance (i.e., Sorensen < 0.7), thus, 466 out of the 490 modeled species were kept for further analysis. The highest species richness (> 75 species) was found in the southwestern and southern regions of the Atlantic Forest. These regions are characterized by high mountainous relief, such as the Serra do Mar Coastal Forests and eastern Araucaria Moist Forest, as well as in the southern region of the Upper Paraná Atlantic Forest (Fig. 1, Suppl. material 1). Species richness gradually decreases towards the east (Upper Paraná Atlantic Forest) and coastal regions of the northeast (Pernambuco Coastal Forests). The northeastern coastal strip showed lower richness (25–50 species). In addition to the northeastern coastal strip, parts of the interior, especially drier areas such as the Caatinga and inland forests of Bahia, also showed low richness (< 25 species; Fig. 1, Suppl. material 1).

Regarding the relationship between species richness and environmental heterogeneity, we found that the variable with the greatest importance (i.e., explained variance) for the Atlantic Forest was temperature seasonality (87.44), followed by topographic (25.7) and geomorphic (23.6) heterogeneity, which each demonstrated similar levels of importance. Stream density had low explanatory power (Fig. 2). Temperature seasonality, topographic heterogeneity, and geomorphic heterogeneity had a positive relationship with species richness, and this relationship was strongest for the first of these variables (Fig. 2, Suppl. material 1: fig. S5). When ecoregion is used in the model, it is the variable with the highest variable importance (56.9), followed by temperature seasonality and topographic and geomorphic heterogeneity (Suppl. material 1: fig. S6).

When we explored the relationship between environmental heterogeneity and bee richness separately for each ecoregion, we found that the influence of environmental heterogeneity on patterns of bee species richness varied among ecoregions. Thus, temperature seasonality was the most important variable in nine of the 11 ecoregions (Fig. 3), with the highest importance values emerging for the Cerrado (97.38), Uruguayan Savanna (92.17), and Campos Rupestres Montane Savanna (77.69). Topographic heterogeneity was a prominent predictor of bee species richness in the Uruguayan Savanna (39.09), Upper Paraná Atlantic Forest (33.87), and Bahia Coastal Forest ecoregions (29.93). Geomorphic heterogeneity and stream density were important variables in the Caatinga (16.29 and 24.30, respectively), Cerrado (41.96, 6.16), and Bahia Coastal Forests (11.21, 0.93). Interestingly, the Serra do Mar Coastal Forests was the ecoregion with the highest species richness, yet also the ecoregion in which the environmental heterogeneity variables explained the least variance in bee species richness (Figs 1, 3).

In addition to varying in relative importance, the direction of the relationship between different aspects of environmental heterogeneity and bee species richness varied among the ecoregions (Fig. 4; Suppl. material 1: fig. S7). While temperature seasonality demonstrated a positive relationship with bee species richness in 5 of the 11 ecoregions (Upper Paraná Atlantic Forests, Bahia Interior Forests, Brazilian Atlantic Dry Forests, Cerrado, and Campos Rupestres Montane Savanna) we observed a negative association between the two variables in 6 ecoregions (Araucaria Moist Forests, Caatinga, Pernambuco Coastal Forests, Serra do Mar Coastal Forests). Topographic heterogeneity and species richness had a positive relationship in 7 ecoregions (the Upper Paraná Atlantic Forest, Bahia Coastal Forests, Brazilian Atlantic Dry Forests, Caatinga, Cerrado, and the Uruguayan Savanna) and a negative relationship in the Campos Rupestres Montane Savanna and Serra do Mar Coastal Forests. Finally, we found that geomorphic heterogeneity showed a positive relationship with species richness in the Bahia Interior Forests, Caatinga, and Campos Rupestres Montane Savanna, but a negative association in the Araucaria Moist Forests, Bahia Coastal Forests, Pernambuco Coastal Forests, and Uruguayan Savanna (Fig. 4; Suppl. material 1: fig. S7). Our analysis of potential circularity between the richness map and the environmental heterogeneity variables revealed that only 27 (5.8%) species had a high correlation between species suitability and temperature seasonality (Suppl. material 1: fig. S8). The richness pattern of these particular species was slightly higher in the southwest region of the study area, but they represented only a small portion of the total richness in that region (Suppl. material 1: fig. S9). This further suggests that, if present, the potential circularity from using similar variables in the distribution and richness modeling should be minimal.

Figure 1.

Species richness patterns of native Atlantic Forest bees throughout the biome and ecoregions. White lines denote the boundaries of the ecoregions.

Figure 2.

Importance of environmental heterogeneity variables in explaining bee species richness and the relationship between these variables and species richness of native bees in the Atlantic Forest. See Suppl. material 1: fig. S5 for results in original variable scales.

Figure 3.

Importance of environmental heterogeneity variables in explaining bee richness in different ecoregions of the Atlantic Forest.

Figure 4.

Relationship between environmental variables and species richness in Atlantic Forest ecoregions. See Suppl. material 1: fig. S7 for results in their original scales.

Discussion

In this study, we sought to describe the patterns of bee diversity in the Atlantic Forest and to investigate the influence of different environmental heterogeneity variables on species richness, both for the entire biome and each of its ecoregions. We found that bee richness was higher in the southern and southwestern regions of the Atlantic Forest and that temperature seasonality and topographic heterogeneity together partly explain these patterns. However, we found that the relationship between species richness and each of the features of environmental heterogeneity that we explored varied among ecoregions. This nonstationarity indicates that the relationship between environmental heterogeneity and bee richness is not spatially homogenous and that some environmental features may be more important for determining species richness in some regions than others (Foody 2004; Eme et al. 2015). Changes in the relationship between biodiversity features and predictor variables at different spatial scales are frequent in spatial ecology (Wang et al. 2012; Eme et al. 2015; Yeager et al. 2017). Understanding how biodiversity relates to the environment at different scales (e.g., biome, ecoregion, and sub-ecoregions) can reveal underlying mechanisms that drive biodiversity patterns in various landscapes. This reinforces the need to complement global analyses of species richness patterns with ecoregion- or landscape-specific assessments to understand richness patterns more deeply (Wang et al. 2012).

Patterns of estimated bee species richness were consistent with patterns observed for other taxonomic groups. For example, the regional hotspots of bee richness (i.e., the southwestern region of the Atlantic Forest) coincide with high concentrations of bird, mammal, and amphibian species (Jenkins et al. 2015); phylogenetic diversity of Opiliones (Arachnida) (Nogueira et al. 2019); and endemic species of amphibians, reptiles, birds, and mammals (Brown et al. 2020; Figueiredo et al. 2021). The general pattern of bee richness also coincides with that of woody plants (Lira-Noriega et al. 2017; Zwiener et al. 2017). Interpreting bee richness patterns demands consideration of the specific geographic and ecological context, especially in transitional zones. We found contrasting scenarios like disjunct Cerrado fragments within the Atlantic Forest (e.g., near to north Araucaria Moist Forest) supporting substantial bee richness, while Atlantic Forest within the Cerrado (e.g., upper portion of the Upper Paraná Atlantic Forest) showed reduced richness. These patterns suggest that geographic proximity to the Atlantic Forest core may influence local species composition. Nonetheless, richness pattern for the Cerrado and Caatinga ecoregions should be strictly limited to the context of this research. Extrapolation to the totality of these ecoregions is inappropriate because our analysis encompassed only portions of these vast areas that were merged with other smaller ecoregions within our study biome (Suppl. material 1: table S1).

We found that temperature seasonality was the most important feature of environmental heterogeneity for bee species richness across the Atlantic Forest biome and within some of its ecoregions. Surprisingly, for most ecoregions (7 of 11), temperature seasonality had a positive relationship with bee species richness. However, this relationship has not been commonly observed in other organisms. For bats and marsupials in the Atlantic Forest, temperature seasonality is negatively associated with species richness (Stevens 2013; Weber and Cáceres 2018; Delciellos et al. 2022), as well as endogenous fungi from Panama (Oita et al. 2021). Nevertheless, previous research has shown that global-scale plant richness is positively associated with temperature seasonality (Tietje et al. 2022). A probable explanation for the positive relationship between temperature seasonality and species richness is that sites with greater temperature seasonality throughout the year can avoid competitive exclusion and harbor a greater number of species niches in different seasons (Tonkin et al. 2017). Seasonal variation is directly linked to plant phenology, affecting flowering and, consequently, resource availability (Escobedo-Kenefic et al. 2020), which influences pollinator diversity.

Topographic and geomorphic heterogeneity showed positive relationships with bee species richness for the entire Atlantic Forest biome as well as several ecoregions in which these environmental features demonstrated high explanatory power. Topographic heterogeneity is important for maintaining biodiversity in the Atlantic Forest (Delciellos et al. 2022). Topographic and geomorphic heterogeneities are related to terrain complexity. As both variables increase, a diverse set of microclimates, vegetation types, and nesting and trophic opportunities for bees are expected (Tukiainen et al. 2019; Doherty et al. 2021; Cavigliasso et al. 2022; Pardee et al. 2023). Steep slopes, valleys, and varying elevations contribute to a mosaic of habitats, each supporting different plant communities, as is the case in the southwest Atlantic Forest, which has the greatest richness of tree species and significant changes in composition. (Zwiener et al. 2017; Rodrigues et al. 2019). Thus, a high diversity of plants can provide various essential food resources for bees (Cavigliasso et al. 2022; Dzekashu et al. 2022; Felderhoff et al. 2023). In addition, variations in microclimates created by topography can promote the coexistence of species with different environmental needs, resulting in greater ecological complexity in areas with more rugged terrain (Muscarella et al. 2020).

Stream density is a relatively understudied variable compared to other metrics of environmental heterogeneity, such as those related to topography or climate. Although water is a fundamental resource for bees (Rankin et al. 2020; Rosa-Fontana et al. 2020; McCune et al. 2021), we found a weak relationship between bee richness and stream density. One likely explanation is that, unlike other organisms, their main needs are the availability of floral resources (Vaudo et al. 2015) and nesting sites (Requier and Leonhardt 2020). However, certain species of bees use water to build nests or regulate temperature within their colonies (Stabentheiner et al. 2010; Stabentheiner et al. 2021). In tropical forests, such as the Atlantic Forest, humidity and high rainfall increase the abundance of floral resources, correlating with greater bee activity, indicating that humidity levels increase foraging opportunities (Silva et al. 2011; Marques et al. 2018). Another probable explanation for this finding is related to the nature of the stream density variable, as it represents potential watercourses, i.e., watercourses that may not exist permanently. Thus, these variables do not distinguish intermittent watercourses from permanent watercourses, especially in the drier regions of the Atlantic Forest (i.e., northeastern Brazil).

Despite the importance of seasonal temperature, topographic, and geomorphic heterogeneity in the Atlantic Forest biome and some ecoregions, these variables demonstrated relatively low explanatory power in ecoregions with the highest species richness, such as the Serra do Mar Coastal Forests and Araucaria Moist Forests. Previous research has shown that the distribution of species richness in areas characterized by especially high diversity cannot be explained by environmental heterogeneity alone (Chang et al. 2023). This finding emphasizes the importance of variables or mechanisms beyond those explained by environmental diversity in determining the diversity patterns of these ecoregions. For example, the Serra do Mar Coastal Forests and Araucaria Moist Forests may harbor high historical climatic stability, promoting favorable conditions for the survival of older species lineages and the emergence of new ones (speciation) (Carnaval et al. 2014; Brown et al. 2020). Globally, solar radiation and precipitation are important variables that explain bee diversity patterns (Orr et al. 2021), making them interesting to be explored at smaller spatial scales Another crucial factor could be the quality and availability of resources, such as floral sources for bees throughout the year, which are essential for maintaining bee diversity (Vaudo et al. 2015; Flo et al. 2018; Klaus et al. 2021). Given the high elevational gradient of the Serra do Mar coastal forest, the low explanatory power in this ecoregion could be related to the spatial resolution of SDM and variables, which may have been too coarse to capture the real variability of the region. Further studies are required to explore other drivers of bee species richness in these regions.

Environmental heterogeneity is increasingly being recognized as an important component in maintaining biodiversity (Knudson et al. 2018; Gordon et al. 2022). Conservation efforts in regions with high environmental heterogeneity have been emphasized to enhance biodiversity conservation because environmental heterogeneity can serve as a proxy for biodiversity (Beier and De Albuquerque 2015; Gordon et al. 2022; Maliniemi et al. 2024). Our results indicated that different components of environmental heterogeneity (temperature seasonality, topographic, and geomorphic heterogeneity) play important roles in determining bee species richness in the Atlantic Forest. Because the relationship between environmental heterogeneity and bee richness was specific to each ecoregion, and each ecoregion likely hosts different species compositions, ecoregion-specific conservation approaches may promote bee conservation throughout the Atlantic Forest. In regions where temperature seasonality was the most important predictor of richness, such as the Cerrado and the Uruguayan Savanna, promoting the conservation of bee habitats that capture climatic gradients may be fundamental for the persistence of local communities. Interestingly, the Serra do Mar Coastal Forests were one of the ecoregions with the highest biodiversity of bees. This ecoregion hosts one of the largest forest remnants of the Atlantic Forest and has been highlighted as a region to expand the current protected areas for amphibians, birds, and plants (Campos et al. 2017; Zwiener et al. 2017; Vale et al. 2018). Because of the large deforestation suffered in the Atlantic Forest, it is necessary to restore areas in ecoregions, such as the Araucaria Moist Forest, Upper Paraná Atlantic Forests, and Cerrado, which have a high biodiversity of bees and other organisms (Zwiener et al. 2017; Crouzeilles et al. 2020; Velazco et al. 2023). Further research could evaluate the contribution of anthropic land cover to the loss of species range and identify priority areas for bee conservation. The limitations of this study are largely related to the intrinsic complexity of the bee group and the ecological variation across bee species in the Atlantic Forest. These species cover a range of different life histories and habitat requirements, making the standardization of occurrence data and generalizability of results particularly challenging. Obtaining occurrence data for bees, and arthropods more broadly, is especially difficult in areas that are not easily accessible. The lack of data for these organisms makes the estimation of species distributions impossible in some cases, which may influence the resulting richness estimates.

Conclusion

Bee species richness was highest in the Serra do Mar Coastal Forests and Araucaria Moist Forest ecoregions, while the coastal regions of the northeast (e.g., the Pernambuco Coastal Forests) and drier inland areas (e.g., the Caatinga and Bahia Inland Forests) showed the lowest diversity. Temperature seasonality emerged as the most important variable explaining species richness, showing a positive relationship with bee richness in most ecoregions. Topographic and geomorphic heterogeneities also play an important role in determining bee richness patterns in some parts of the Atlantic Forest. Moreover, stream density was poorly correlated with richness patterns. We found that the relationship between environmental heterogeneity variables and bee species richness varied substantially between ecoregions. This finding reinforces the need to study diversity patterns at multiple spatial scales to highlight the nuanced drivers of species diversity within each ecoregion. Interestingly, in ecoregions with the highest bee species richness, aspects of environmental heterogeneity showed low explanatory power, suggesting that other factors are the primary determinants of local biodiversity.

Acknowledgements

We thank B. Rose for reading and editing the first draft of the manuscript. AIA and ACOJ thank CAPES-DS for their master’s degree scholarship. FV Thanks INECOL and CONAHCyT-Mexico for their support. SJEV thanks i) the Center for Open Geographical Science (COGS), Department of Geography, at San Diego State University for the research support, ii) Atlantic Forest Biodiversity Observatory (Instituto de Biologia Subtropical, UNaM CONICET) for computational resources, and iii) funding from FONCyT, Agencia de Ciencia y Tecnología, Ministerio de Ciencia y Tecnología de Argentina (PICTO-2022-10-00097).

Author contributions

Conceptualization (AIA, SJEV), Methodology (AIA, SJEV), Formal analysis (AIA, ACOJ, SJEV), Data curation (AIA, SJEV), Writing – Original Draft (AIA, SJEV), Writing – Review & Editing (AIA, ACOJ, FV, SJEV).

Data accessibility statement

Species occurrence dataset and predictor variables used to construct species distribution, and raster with environmental heterogeneity variables can be accessed at: https://doi.org/10.6084/m9.figshare.28680413.

References

Amatulli G, Garcia Marquez J, Sethi T, Kiesel J, Grigoropoulou A, Üblacker MM, Shen LQ, Domisch S (2022) Hydrography90m: a new high-resolution global hydrographic dataset. Earth System Science Data 14: 4525–4550. https://doi.org/10.5194/essd-14-4525-2022

Amatulli G, McInerney D, Sethi T, Strobl P, Domisch S (2020) Geomorpho90m, empirical evaluation and accuracy assessment of global high-resolution geomorphometric layers. Scientific Data 7: 1–18. https://doi.org/10.1038/s41597-020-0479-6

Andrella GC, Koch I, Velazco SJE (2023) Considering spatial constraints to identify areas for new species sampling: A species-specific prioritization approach. Biological Conservation 288: 110379. https://doi.org/10.1016/j.biocon.2023.110379

Baker DJ, Maclean IMD, Goodall M, Gaston KJ (2022) Correlations between spatial sampling biases and environmental niches affect species distribution models. Global Ecology and Biogeography 31: 1038–1050. https://doi.org/10.1111/geb.13491

Batista CB, Lima IP, Lima MR (2021) Beta diversity patterns of bats in the Atlantic Forest: How does the scale of analysis affect the importance of spatial and environmental factors? Journal of Biogeography 48: 1–10. https://doi.org/10.1111/jbi.13928

Beier P, Albuquerque FS (2015) Environmental diversity as a surrogate for species representation: Environmental Diversity as a Surrogate. Conservation Biology 29: 1401–1410. https://doi.org/10.1111/cobi.12495

Breiner FT, Guisan A, Bergamini A Nobis MP (2015) Overcoming limitations of modelling rare species by using ensembles of small models. Methods in Ecology and Evolution 6: 1210–1218. https://doi.org/10.1111/2041-210X.12403

Brown JL, Paz A, Reginato M, Renata CA, Assis C, Lyra M, Caddah MK, Aguirre-Santoro J, d’Horta F, Raposo do Amaral F, Goldenberg R, Silva-Brandão KL, Freitas AVL, Rodrigues MT, Michelangeli FA, Miyaki CY, Carnaval AC (2020) Seeing the forest through many trees: Multi-taxon patterns of phylogenetic diversity in the Atlantic Forest hotspot. Diversity and Distributions 26: 1160–1176. https://doi.org/10.1111/ddi.13116

Brown SC, Wigley TML, Otto-Bliesner BL, Rahbek C, Fordham DA (2020) Persistent Quaternary climate refugia are hospices for biodiversity in the Anthropocene. Nature Climate Change 10: 244–248. https://doi.org/10.1038/s41558-019-0682-7

Bukowski B, Ratnasingham S, Hanisch PE, Hebert PDN, Perez K, deWaard J, Tubaro PL, Lijtmaer DA (2022) DNA barcodes reveal striking arthropod diversity and unveil seasonal patterns of variation in the southern Atlantic Forest. PLoS ONE 17: e0267390. https://doi.org/10.1371/journal.pone.0267390

Burdine JD, McCluney KE (2019) Differential sensitivity of bees to urbanization-driven changes in body temperature and water content. Scientific Reports 9: 1643. https://doi.org/10.1038/s41598-018-38338-0

Campos FS, Lourenço-de-Moraes R, Llorente GA, Solé M (2017) Cost-effective conservation of amphibian ecology and evolution. Science Advances 3: e1602929. https://doi.org/10.1126/sciadv.1602929

Carnaval AC, Waltari E, Rodrigues MT, Rosauer D, VanDerWal J, Damasceno R, Prates I, Strangas M, Spanos Z, Rivera D, Pie MR, Firkowski CR, Bornschein MR, Ribeiro LF, Moritz C (2014) Prediction of phylogeographic endemism in an environmentally complex biome. Proceedings of the Royal Society B: Biological Sciences 281: 20141461. https://doi.org/10.1098/rspb.2014.1461

Carpenter G, Gillison AN, Winter J (1993) DOMAIN: a flexible modelling procedure for mapping potential distributions of plants and animals. Biodiversity and Conservation 2: 667–680. https://doi.org/10.1007/BF00051966

Carvalho CS, Martello F, Galetti M, Pinto F, Francisco MR, Silveira LF, Galetti PM (2021) Environmental heterogeneity and sampling relevance areas in an Atlantic forest endemism region. Perspectives in Ecology and Conservation 19: 311–318. https://doi.org/10.1016/j.pecon.2021.05.001

Cavigliasso P, Phifer CC, Knowlton JL, Licata JA, Flaspohler DJ, Webster CR, Chacoff NP (2022) Influence of landscape composition on wild bee communities: Effects of functional landscape heterogeneity. Agriculture, Ecosystems & Environment 340: 108150. https://doi.org/10.1016/j.agee.2022.108150

Chang Y, Gelwick K, Willett SD, Shen X, Albouy C, Luo A, Wang Z, Zimmermann NE, Pellissier L (2023) Phytodiversity is associated with habitat heterogeneity from Eurasia to the Hengduan Mountains. New Phytologist 240: 1647–1658. https://doi.org/10.1111/nph.19206

Classen A, Peters MK, Kindeketa WJ, Appelhans T, Eardley CD, Gikungu MW, Hemp A, Nauss T, Steffan-Dewenter I (2015) Temperature versus resource constraints: which factors determine bee diversity on Mount Kilimanjaro, Tanzania? Global Ecology and Biogeography 24: 642–652. https://doi.org/10.1111/geb.12286

Cortopassi-Laurino M, Velthuis H, Nogueira-Neto P (2007) Diversity of stingless bees from the Amazon forest in Xapuri (Acre). Proceedings of the Netherlands Entomological Society Meeting 18: 105–114.

Crouzeilles R, Beyer HL, Monteiro LM, Feltran-Barbieri R, Pessôa ACM, Barros FSM, Lindenmayer DB, Lino EDSM, Grelle CEV, Chazdon RL, Matsumoto M, Rosa M, Latawiec AE, Strassburg BBN (2020) Achieving cost-effective landscape-scale forest restoration through targeted natural regeneration. Conservation Letters 13: e12709. https://doi.org/10.1111/conl.12709

De Marco P, Nóbrega CC (2018) Evaluating collinearity effects on species distribution models: An approach based on virtual species simulation. PLoS ONE 13: e0202403. https://doi.org/10.1371/journal.pone.0202403

Delciellos AC, Prevedello JA, Figueiredo MSL, Weber MM, Lorini ML (2022) Species Richness and Endemism of Marsupials in the Atlantic Forest: Spatial Patterns and Drivers. In: Cáceres N, Dickman CR (Eds) American and Australasian Marsupials. Springer International Publishing, Cham, 1–23. https://doi.org/10.1007/978-3-030-88800-8_17-1

Dinerstein E, Olson D, Joshi A, Vynne C, Burgess ND, Wikramanayake E, Hahn N, Palminteri S, Hedao P, Noss R, Hansen M, Locke H, Ellis EC, Jones B, Barber CV, Hayes R, Kormos C, Martin V, Crist E, Sechrest W, Price L, Baillie JEM, Weeden D, Suckling K, Davis C, Sizer N, Moore R, Thau D, Birch T, Potapov P, Turubanova S, Tyukavina A, Souza N, Pintea L, Brito JC, Llewellyn OA, Miller AG, Patzelt A, Ghazanfar SA, Timberlake J, Klöser H, Shennan-Farpón Y, Kindt R, Lillesø JPB, van Breugel P, Graudal L, Voge M, Al-Shammari KF, Saleem M (2017) An ecoregion-based approach to protecting half the terrestrial realm. BioScience 67: 534–545. https://doi.org/10.1093/biosci/bix014

Doherty KD, Kuhlman MP, Durham RA, Ramsey PW, Mummey DL (2021) Fine-grained topographic diversity data improve site prioritization outcomes for bees. Ecological Indicators 132: 108315. https://doi.org/10.1016/j.ecolind.2021.108315

Domisch S, Friedrichs M, Hein T, Borgwardt F, Wetzig A, Jähnig SC, Langhans SD (2019) Spatially explicit species distribution models: A missed opportunity in conservation planning? Diversity and Distributions 25: 758–769. https://doi.org/10.1111/ddi.12891

Dool SE, Picker MD, Eberhard MJB (2022) Limited dispersal and local adaptation promote allopatric speciation in a biodiversity hotspot. Molecular Ecology 31: 279–295. https://doi.org/10.1111/mec.16219

Dorey JB, Fischer EE, Chesshire PR, Nava-Bolaños A, O’Reilly RL, Bossert S, Collins SM, Lichtenberg EM, Tucker EM, Smith-Pardo A, Falcon-Brindis A, Guevara DA, Ribeiro B, De Pedro D, Pickering J, Hung KLJ, Parys KA, McCabe LM, Rogan MS, Minckley RL, Velazco SJE, Griswold T, Zarrillo TA, Jetz W, Sica YV, Orr MC, Guzman LM, Ascher JS, Hughes AC, Cobb NS (2023) A globally synthesised and flagged bee occurrence dataset and cleaning workflow. Scientific Data 10: 747. https://doi.org/10.1038/s41597-023-02626-w

Dufour A, Gadallah F, Wagner HH, Guisan A, Buttler A (2006) Plant species richness and environmental heterogeneity in a mountain landscape: effects of variability and spatial configuration. Ecography 29: 573–584. https://doi.org/10.1111/j.0906-7590.2006.04605.x

Dzekashu FF, Yusuf AA, Pirk CWW, Steffan-Dewenter I, Lattorff HMG, Peters MK (2022) Floral turnover and climate drive seasonal bee diversity along a tropical elevation gradient. Ecosphere 13: e3964. https://doi.org/10.1002/ecs2.3964

Eme D, Zagmajster M, Fišer C, Galassi D, Marmonier P, Stoch F, Cornu J, Oberdorff T, Malard F (2015) Multi-causality and spatial non-stationarity in the determinants of groundwater crustacean diversity in Europe. Ecography 38: 531–540. https://doi.org/10.1111/ecog.01092

Escobedo-Kenefic N, Landaverde-González P, Theodorou P, Cardona E, Dardón MJ, Martínez O, Domínguez CA (2020) Disentangling the effects of local resources, landscape heterogeneity and climatic seasonality on bee diversity and plant-pollinator networks in tropical highlands. Oecologia 194: 333–344. https://doi.org/10.1007/s00442-020-04715-8

Faoro H, de Souza EM, Pedrosa FO (2015) Brazilian Atlantic Forest Soil Metagenome. In: Highlander SK, Rodriguez-Valera F, White BA (Eds) Encyclopedia of Metagenomics: Environmental Metagenomics. Springer US, Boston, 54–59. https://doi.org/10.1007/978-1-4899-7475-4_781

Feitosa RM, De Castro Morini MS, Martins AC, De Andrade Ribeiro TM, Noll FB, Dos Santos EF, Cancello EM, Constantini JP (2021) Social Insects of the Atlantic Forest. In: Marques MCM, Grelle CEV (Eds) The Atlantic Forest. Springer International Publishing, Cham, 151–183. https://doi.org/10.1007/978-3-030-55322-7_8

Felderhoff J, Gathof AK, Buchholz S, Egerer M (2023) Vegetation complexity and nesting resource availability predict bee diversity and functional traits in community gardens. Ecological Applications 33: e2759. https://doi.org/10.1002/eap.2759

Ferrier S, Guisan A (2006) Spatial modelling of biodiversity at the community level. Journal of applied ecology 43: 393–404. https://doi.org/10.1111/j.1365-2664.2006.01149.x

Figueiredo MDSL, Weber MM, Brasileiro CA, Cerqueira R, Grelle CEV, Jenkins CN, Solidade CV, Thomé MTC, Vale MM, Lorini ML (2021) Tetrapod Diversity in the Atlantic Forest: Maps and Gaps. In: Marques MCM, Grelle CEV (Eds) The Atlantic Forest. Springer International Publishing, Cham, 185–204. https://doi.org/10.1007/978-3-030-55322-7_9

Flo V, Bosch J, Arnan X, Primante C, Martín González AM, Barril-Graells H, Rodrigo A (2018) Yearly fluctuations of flower landscape in a Mediterranean scrubland: Consequences for floral resource availability. PLoS ONE 13: e0191268. https://doi.org/10.1371/journal.pone.0191268

Foody GM (2004) Spatial nonstationarity and scale-dependency in the relationship between species richness and environmental determinants for the sub-Saharan endemic avifauna. Global Ecology and Biogeography 13: 315–320. https://doi.org/10.1111/j.1466-822X.2004.00097.x

Fortin MJ, Dale MRT (2005) Spatial analysis a guide for ecologists. Cambridge University Press, Cambridge, N.Y. https://doi.org/10.1017/CBO9780511542039

Fourcade Y (2021) Fine-tuning niche models matters in invasion ecology. A lesson from the land planarian Obama nungara. Ecological Modelling 457: 109686. https://doi.org/10.1016/j.ecolmodel.2021.109686

Garraffoni ARS, Moura FR, Lourenço AP (2017) Areas of endemism in the Atlantic Forest: quantitative biogeography insights from orchid bees (Apidae: Euglossini). Apidologie 48: 513–522. https://doi.org/10.1007/s13592-017-0494-6

Gordon JE, Bailey JJ, Larwood JG (2022) Conserving nature’s stage provides a foundation for safeguarding both geodiversity and biodiversity in protected and conserved areas. Parks Stewardship Forum 38: 46–55. https://doi.org/10.5070/P538156118

Greenwell BM, Boehmke BC (2020) Variable Importance Plots—An Introduction to the vip Package. The R Journal 12: 1. https://doi.org/10.32614/RJ-2020-013

Guillera-Arroita G, Lahoz-Monfort JJ, Elith J, Gordon A, Kujala H, Lentini PE, McCarthy MA, Tingley R, Wintle BA (2015) Is my species distribution model fit for purpose? Matching data and models to applications. Global Ecology and Biogeography 24: 276–292. https://doi.org/10.1111/geb.12268

Guisan A, Rahbek C (2011) SESAM – a new framework integrating macroecological and species distribution models for predicting spatio-temporal patterns of species assemblages. Journal of Biogeography 38: 1433–1444. https://doi.org/10.1111/j.1365-2699.2011.02550.x

Hoffmeister CH, Ferrari A (2016) Areas of endemism of arthropods in the Atlantic Forest (Brazil): an approach based on a metaconsensus criterion using endemicity analysis. Biological Journal of the Linnean Society 119: 126–144. https://doi.org/10.1111/bij.12802

Jenkins CN, Alves MAS, Uezu A, Vale MM (2015) Patterns of Vertebrate Diversity and Protection in Brazil. PLoS ONE 10: e0145064. https://doi.org/10.1371/journal.pone.0145064

Karger DN, Conrad O, Böhner J, Kawohl T, Kreft H, Soria-Auza RW, Zimmermann NE, Linder HP, Kessler M (2017) Climatologies at high resolution for the earth’s land surface areas. Scientific Data 4: 170122. https://doi.org/10.1038/sdata.2017.122

Keeton WS, Kraft CE, Warren DR (2007) Mature and old-growth riparian forests: structure, dynamics, and effects on Adirondack stream habitats. Ecological Applications 17: 852–868. https://doi.org/10.1890/06-1172

Klaus F, Tscharntke T, Bischoff G, Grass I (2021) Floral resource diversification promotes solitary bee reproduction and may offset insecticide effects – evidence from a semi-field experiment. Ecology Letters 24: 668–675. https://doi.org/10.1111/ele.13683

Klein AM, Vaissière BE, Cane JH, Steffan-Dewenter I, Cunningham SA, Kremen C, Tscharntke T (2006) Importance of pollinators in changing landscapes for world crops. Proceedings of the Royal Society B: Biological Sciences 274: 303–313. https://doi.org/10.1098/rspb.2006.3721

Klein S, Cabirol A, Devaud JM, Barron AB, Lihoreau M (2017) Why bees are so vulnerable to environmental stressors. Trends in Ecology & Evolution 32: 268–278. https://doi.org/10.1016/j.tree.2016.12.009

Knudson C, Kay K, Fisher S (2018) Appraising geodiversity and cultural diversity approaches to building resilience through conservation. Nature Climate Change 8: 678–685. https://doi.org/10.1038/s41558-018-0188-8

Kremen C, Williams NM, Bugg RL, Fay JP, Thorp RW (2004) The area requirements of an ecosystem service: crop pollination by native bee communities in California. Ecology Letters 7: 1109–1119. https://doi.org/10.1111/j.1461-0248.2004.00662.x

Kühnholz S, Seeley TD (1997) The control of water collection in honey bee colonies. Behavioral Ecology and Sociobiology 41: 407–422. https://doi.org/10.1007/s002650050402

Lenth RV (2016) Least-Squares Means: The R Package lsmeans. Journal of Statistical Software 69(1): 1–33. https://doi.org/10.18637/jss.v069.i01

Leroy B, Delsol R, Hugueny B, Meynard CN, Barhoumi C, Barbet-Massin M, Bellard C (2018) Without quality presence–absence data, discrimination metrics such as TSS can be misleading measures of model performance. Journal of Biogeography 45: 1994–2002. https://doi.org/10.1111/jbi.13402

Liu C, Newell G, White M (2016) On the selection of thresholds for predicting species occurrence with presence-only data. Ecology and Evolution 6: 337–348. https://doi.org/10.1002/ece3.1878

Maliniemi T, Tukiainen H, Hjort J, Toivanen M, Vernham G, Bailey JJ, Baines O, Benniston L, Brilha J, Field R, Fox N, Gray M, Grytnes JA, Huusko K, Kemppinen J, Pereira P, Salminen H, Schrodt F, Turner L, Alahuhta J (2024) Too much diversity—Multiple definitions of geodiversity hinder its potential in biodiversity research. Diversity and Distributions 30: e13843. https://doi.org/10.1111/ddi.13843

Marques MF, Deprá MS, Gaglianone MC (2018) Seasonal Variation in Bee-Plant Interactions in an Inselberg in the Atlantic Forest in Southeastern Brazil. Sociobiology 65: 612. https://doi.org/10.13102/sociobiology.v65i4.3473

McCune F, Samson-Robert O, Rondeau S, Chagnon M, Fournier V (2021) Supplying honey bees with waterers: a precautionary measure to reduce exposure to pesticides. Environmental Science and Pollution Research 28: 17573–17586. https://doi.org/10.1007/s11356-020-12147-3

Mendes P, Velazco SJE, Andrade AFA, De Marco P (2020) Dealing with overprediction in species distribution models: How adding distance constraints can improve model accuracy. Ecological Modelling 431: 109180. https://doi.org/10.1016/j.ecolmodel.2020.109180

Michener CD (2007) The Bees of the World. The Johns Hopkins University Press, Baltimore.

Mittermeier RA, Gil PR, Hoffman M, Pilgrim J, Brooks T, Mittermeier CG, Lamoreux J, Fonseca GAB (2005) Hotspots Revisited: Earth’s Biologically Richest and Most Endangered Terrestrial Ecoregions. Conservation International, Mexico City.

Morellato LPC, Haddad CFB (2000) Introduction: The Brazilian atlantic forest. Biotropica 32: 786–792. https://doi.org/10.1111/j.1744-7429.2000.tb00618.x

Muscarella R, Kolyaie S, Morton DC, Zimmerman JK, Uriarte M (2020) Effects of topography on tropical forest structure depend on climate context. Journal of Ecology 108: 145–159. https://doi.org/10.1111/1365-2745.13261

Muylaert RL, Vancine MH, Bernardo R, Oshima JEF, Sobral-Souza T, Tonetti VR, Niebuhr BB, Ribeiro MC (2018) Uma nota sobre os limites territoriais da Mata Atlântica. Oecologia Australis 22: 302–311. https://doi.org/10.4257/oeco.2018.2203.09

Myers N, Mittermeier RA, Mittermeier CG, da Fonseca GAB, Kent J (2000) Biodiversity hotspots for conservation priorities. Nature 403: 853–858. https://doi.org/10.1038/35002501

Nogueira ADA, Bragagnolo C, DaSilva MB, Carvalho LS, Benedetti AR, Pinto-da-Rocha R (2019) Spatial variation in phylogenetic diversity of communities of Atlantic Forest harvestmen (Opiliones, Arachnida). Insect Conservation and Diversity 12: 414–426. https://doi.org/10.1111/icad.12356

O’Donnell MS, Ignizio DA (2012) Bioclimatic predictors for supporting ecological applications in the conterminous United States. Geological Survey 691: 1–10.

Oita S, Ibáñez A, Lutzoni F, Miadlikowska J, Geml J, Lewis LA, Hom EFY, Carbone I, U’Ren JM, Arnold AE (2021) Climate and seasonality drive the richness and composition of tropical fungal endophytes at a landscape scale. Communications Biology 4: 313. https://doi.org/10.1038/s42003-021-01826-7

Ollerton J, Winfree R, Tarrant S (2011) How many flowering plants are pollinated by animals? Oikos 120: 321–326. https://doi.org/10.1111/j.1600-0706.2010.18644.x

Olson DM, Dinerstein E, Wikramanayake ED, Burgess ND, Powell GV, Underwood EC, D’amico JA, Itoua I, Strand HE, Morrison JC (2001) Terrestrial Ecoregions of the World: A New Map of Life on Earth: A new global map of terrestrial ecoregions provides an innovative tool for conserving biodiversity. BioScience 51: 933–938. https://doi.org/10.1641/0006-3568(2001)051[0933:TEOTWA]2.0.CO;2

Orr MC, Hughes AC, Chesters D, Pickering J, Zhu CD, Ascher JS (2021) Global Patterns and Drivers of Bee Distribution. Current Biology 31: 451–458. https://doi.org/10.1016/j.cub.2020.10.053

Pardee GL, Ballare KM, Neff JL, Do LQ, Ojeda D, Bienenstock EJ, Brosi BJ, Grubesic TH, Miller JA, Tong D, Jha S. (2023) Local and landscape factors influence plant-pollinator networks and bee foraging behavior across an urban corridor. Land 12: 362. https://doi.org/10.3390/land12020362

Pebesma E, Bivand R (2023) Spatial Data Science: With Applications in R. Chapman and Hall/CRC, New York. https://doi.org/10.1201/9780429459016

Peterson AT, Soberón J, Pearson RG, Anderson RP, Martínez-Meyer E, Nakamura M, Araújo MB (2011) Ecological niches and geographic distributions. Princeton University Press, Princeton, N.J. https://doi.org/10.1515/9781400840670

Phillips SJ, Anderson RP, Schapire RE (2006) Maximum entropy modeling of species geographic distributions. Ecological Modelling 190: 231–259. https://doi.org/10.1016/j.ecolmodel.2005.03.026

Potts SG, Vulliamy B, Dafni A, Ne’eman G, Willmer P (2003) Linking Bees and Flowers: How Do Floral Communities Structure Pollinator Communities? Ecology 84: 2628–2642. https://doi.org/10.1890/02-0136

Prado DE, Gibbs PE (1993) Patterns of Species Distributions in the Dry Seasonal Forests of South America. Annals of the Missouri Botanical Garden 80: 902–927. https://doi.org/10.2307/2399937

R Core Team (2024) R: A language and environment for statistical computing. R Foundation for Statistical Computing.

Ramos SE, Schiestl FP (2019) Rapid plant evolution driven by the interaction of pollination and herbivory. Science 364: 193–196. https://doi.org/10.1126/science.aav6962

Rankin EEW, Barney SK, Lozano GE (2020) Reduced water negatively impacts social bee survival and productivity via shifts in floral nutrition. Journal of Insect Science 20: 1–8. https://doi.org/10.1093/jisesa/ieaa114

Requier F, Leonhardt SD (2020) Beyond flowers: including non-floral resources in bee conservation schemes. Journal of Insect Conservation 24: 5–16. https://doi.org/10.1007/s10841-019-00206-1

Ribeiro BR, Velazco SJE, Guidoni-Martins K, Tessarolo G, Jardim L, Bachman SP, Loyola R (2022) bdc : A toolkit for standardizing, integrating and cleaning biodiversity data. Methods in Ecology and Evolution 13: 1421–1428. https://doi.org/10.1111/2041-210X.13868

Ribeiro MC, Metzger JP, Martensen AC, Ponzoni FJ, Hirota MM (2009) The Brazilian Atlantic Forest: How much is left, and how is the remaining forest distributed? Implications for conservation. Biological Conservation 142: 1141–1153. https://doi.org/10.1016/j.biocon.2009.02.021

Ricketts TH, Regetz J, Steffan-Dewenter I, Cunningham SA, Kremen C, Bogdanski A, Gemmill-Herren B, Greenleaf SS, Klein AM, Mayfield MM, Morandin LA, Ochieng’ A, Viana BF (2008) Landscape effects on crop pollination services: are there general patterns? Ecology Letters 11: 499–515. https://doi.org/10.1111/j.1461-0248.2008.01157.x

Ricklefs RE (2004) A comprehensive framework for global patterns in biodiversity. Ecology Letters 7: 1–15. https://doi.org/10.1046/j.1461-0248.2003.00554.x

Rodrigues AC, Villa PM, Neri AV (2019) Fine-scale topography shape richness, community composition, stem and biomass hyperdominant species in Brazilian Atlantic forest. Ecological Indicators 102: 208–217. https://doi.org/10.1016/j.ecolind.2019.02.033

Rogers SR, Tarpy DR, Burrack HJ (2014) Bee Species Diversity Enhances Productivity and Stability in a Perennial Crop. PLoS ONE 9: e97307. https://doi.org/10.1371/journal.pone.0097307

Rollinson CR, Finley AO, Alexander MR, Banerjee S, Dixon Hamil K-A, Koenig LE, Locke DH, DeMarche ML, Tingley MW, Wheeler K, Youngflesh C, Zipkin EF (2021) Working across space and time: nonstationarity in ecological research and application. Frontiers in Ecology and the Environment 19: 66–72. https://doi.org/10.1002/fee.2298

Rosa-Fontana AS, Dorigo AS, Soares-Lima HM, Nocelli RCF, Malaspina O (2020) Is the water supply a key factor in stingless bees’ intoxication? Journal of Insect Science 20: 1–4. https://doi.org/10.1093/jisesa/ieaa127

Rose MB, Velazco SJE, Regan HM, Flint AL, Flint LE, Thorne JH, Franklin J (2024) Uncertainty in consensus predictions of plant species’ vulnerability to climate change. Diversity and distributions 30: e13898. https://doi.org/10.1111/ddi.13898

Rubene D, Schroeder M, Ranius T (2015) Diversity patterns of wild bees and wasps in managed boreal forests: Effects of spatial structure, local habitat and surrounding landscape. Biological Conservation 184: 201–208. https://doi.org/10.1016/j.biocon.2015.01.029

Rushton SP, Ormerod SJ, Kerby G (2004) New paradigms for modelling species distributions? Journal of Applied Ecology 41: 193–200. https://doi.org/10.1111/j.0021-8901.2004.00903.x

Silva MDE, Ramalho M, Rosa JF (2011) Por que Melipona scutellaris (Hymenoptera, Apidae) forrageia sob alta umidade relativa do ar? Iheringia. Série Zoologia 101: 131–137. https://doi.org/10.1590/S0073-47212011000100019

Souza Y, Gonçalves F, Lautenschlager L, Akkawi P, Mendes C, Carvalho MM, Bovendorp RS, Fernandes-Ferreira H, Rosa C, Graipel ME, Peroni N, Cherem JJ, Bogoni JA, Brocardo CR, Miranda J, Zago da Silva L, Melo G, Cáceres N, Sponchiado J, Ribeiro MC, Galetti M (2019) ATLANTIC MAMMALS: a data set of assemblages of medium- and large-sized mammals of the Atlantic Forest of South America. Ecology 100: e02785. https://doi.org/10.1002/ecy.2785

Stabentheiner A, Kovac H, Brodschneider R (2010) Honeybee Colony Thermoregulation – Regulatory Mechanisms and Contribution of Individuals in Dependence on Age, Location and Thermal Stress. PLoS ONE 5: e8967. https://doi.org/10.1371/journal.pone.0008967

Stabentheiner A, Kovac H, Mandl M, Käfer H (2021) Coping with the cold and fighting the heat: thermal homeostasis of a superorganism, the honeybee colony. Journal of Comparative Physiology A 207: 337–351. https://doi.org/10.1007/s00359-021-01464-8

Stehmann JR, Forzza RC, Salino A, Sobral M, Costa DP, Kamino LHY (2009) Plantas da Floresta Atlântica. Jardim Botânico do Rio de Janeiro, Rio de Janeiro.

Stein A, Gerstner K, Kreft H (2014) Environmental heterogeneity as a universal driver of species richness across taxa, biomes and spatial scales. Ecology Letters 17: 866–880. https://doi.org/10.1111/ele.12277

Stein A, Kreft H (2015) Terminology and quantification of environmental heterogeneity in species-richness research. Biological Reviews 90: 815–836. https://doi.org/10.1111/brv.12135

Stevens RD (2013) Gradients of Bat Diversity in Atlantic Forest of South America: Environmental Seasonality, Sampling Effort and Spatial Autocorrelation. Biotropica 45: 764–770. https://doi.org/10.1111/btp.12056

Stolar J, Nielsen SE (2015) Accounting for spatially biased sampling effort in presence-only species distribution modelling. Diversity and Distributions 21: 595–608. https://doi.org/10.1111/ddi.12279

Swets JA (1988) Measuring the accuracy of diagnostic systems. Science 240: 1285–1293. https://doi.org/10.1126/science.3287615

Tiefelsdorf M, Griffith DA (2007) Semiparametric Filtering of Spatial Autocorrelation: The Eigenvector Approach. Environment and Planning A: Economy and Space 39: 1193–1221. https://doi.org/10.1068/a37378

Tietje M, Antonelli A, Baker WJ, Govaerts R, Smith SA, Eiserhardt WL (2022) Global variation in diversification rate and species richness are unlinked in plants. Proceedings of the National Academy of Sciences 119: e2120662119. https://doi.org/10.1073/pnas.2120662119

Tonkin JD, Bogan MT, Bonada N, Rios-Touma B, Lytle DA (2017) Seasonality and predictability shape temporal species diversity. Ecology 98: 1201–1216. https://doi.org/10.1002/ecy.1761

Tukiainen H, Alahuhta J, Field R, Ala-Hulkko T, Lampinen R, Hjort J (2017) Spatial relationship between biodiversity and geodiversity across a gradient of land-use intensity in high-latitude landscapes. Landscape Ecology 32: 1049–1063. https://doi.org/10.1007/s10980-017-0508-9

Tukiainen H, Kiuttu M, Kalliola R, Alahuhta J, Hjort J (2019) Landforms contribute to plant biodiversity at alpha, beta and gamma levels. Journal of Biogeography 46: 1699–1710. https://doi.org/10.1111/jbi.13569

Vale MM, Souza TV, Alves MAS, Crouzeilles R (2018) Planning protected areas network that are relevant today and under future climate change is possible: the case of Atlantic Forest endemic birds. PeerJ 6: e4689. https://doi.org/10.7717/peerj.4689

Varela S, Anderson RP, García-Valdés R, Fernández-González F (2014) Environmental filters reduce the effects of sampling bias and improve predictions of ecological niche models. Ecography 37: 1084–1091. https://doi.org/10.1111/j.1600-0587.2013.00441.x

Vaudo AD, Tooker JF, Grozinger CM, Patch HM (2015) Bee nutrition and floral resource restoration. Current Opinion in Insect Science 10: 133–141. https://doi.org/10.1016/j.cois.2015.05.008

Velazco SJE, Ribeiro BR, Laureto LMO, De Marco Jr P (2020) Overprediction of species distribution models in conservation planning: A still neglected issue with strong effects. Biological Conservation 252: 108822. https://doi.org/10.1016/j.biocon.2020.108822

Velazco SJE, Rose MB, Andrade AFA, Minoli I, Franklin J (2022) flexsdm: An R package for supporting a comprehensive and flexible species distribution modelling workflow. Methods in Ecology and Evolution 13: 1661–1669. https://doi.org/10.1111/2041-210X.13874

Velazco SJE, Svenning JC, Ribeiro BR, Laureto LMO (2021) On opportunities and threats to conserve the phylogenetic diversity of Neotropical palms. Diversity and Distributions 27: 512–523. https://doi.org/10.1111/ddi.13215

Velazco SJE, Villalobos F, Galvão F, De Marco Jr P (2023) Transboundary conservation opportunities for Cerrado’s plant species. Biological Conservation 284: 110194. https://doi.org/10.1016/j.biocon.2023.110194

Wan JZ, Wang CJ, Marquet PA (2023) Environmental heterogeneity as a driver of terrestrial biodiversity on a global scale. Progress in Physical Geography: Earth and Environment 47: 912–930. https://doi.org/10.1177/03091333231189045

Wang Z, Rahbek C, Fang J (2012) Effects of geographical extent on the determinants of woody plant diversity. Ecography 35: 1160–1167. https://doi.org/10.1111/j.1600-0587.2012.07786.x

Weber MDM, Cáceres NC (2018) Spatial variation in bat species richness in a forest-grassland transitional region of southeastern South America. Oecologia Australis 22: 156–167. https://doi.org/10.4257/oeco.2018.2202.05

Whalen MA, Aquilino KM, Stachowicz JJ (2016) Grazer diversity interacts with biogenic habitat heterogeneity to accelerate intertidal algal succession. Ecology 97: 2136–2146. https://doi.org/10.1890/15-1633.1

Xu W, Ci X, Song C, He T, Zhang W, Li Q, Li J (2016) Soil phosphorus heterogeneity promotes tree species diversity and phylogenetic clustering in a tropical seasonal rainforest. Ecology and Evolution 6: 8719–8726. https://doi.org/10.1002/ece3.2529

Yeager LA, Deith MCM, McPherson JM, Williams ID, Baum JK (2017) Scale dependence of environmental controls on the functional diversity of coral reef fish communities. Global Ecology and Biogeography 26: 1177–1189. https://doi.org/10.1111/geb.12628

Zizka A, Silvestro D, Andermann T, Azevedo J, Duarte Ritter C, Edler D, Farooq H, Herdean A, Ariza M, Scharn R, Svantesson S, Wengström N, Zizka V, Antonelli A (2019) CoordinateCleaner: Standardized cleaning of occurrence records from biological collection databases. Methods in Ecology and Evolution 10: 744–751. https://doi.org/10.1111/2041-210X.13152

Zwiener VP, Lira-Noriega A, Grady CJ, Padial AA, Vitule JRS (2017) Climate change as a driver of biotic homogenization of woody plants in the Atlantic Forest. Global Ecology and Biogeography 27: 298–309. https://doi.org/10.1111/geb.12695

Zwiener VP, Padial AA, Marques MCM, Faleiro FV, Loyola R, Peterson AT (2017) Planning for conservation and restoration under climate and land use change in the Brazilian Atlantic Forest. Diversity and Distributions 23: 955–966.https://doi.org/10.1111/ddi.12588

Supplementary material

Supplementary material 1

Figure S1: Map of Atlantic Forest ecoregions; Figure S2: Correlogram of the original predictor variables based on the Pearson correlation matrix; Figure S3: Geographical pattern of environmental heterogeneity variables used to relate with bee richness of the Atlantic Forest; Figure S4: Performance of ESM (esm_) and consensus models (median) based on Sorensen and IMAE metrics; Figure S5: Relationship between different environmental heterogeneity variables in their original scales and species richness of native bees in the Atlantic Forest; Figure S6: Importance of ecoregion and different environmental heterogeneity variables in explaining species richness of native bees in the Atlantic Forest; Figure S7: Relationship between environmental variables in their original scales and species richness in Atlantic Forest ecoregions; Figure S8: Pearson correlation between species suitability and predictor variables used to explore the relationship between species richness and environmental heterogeneity. Each point represents a species. Red-dashed lines represent correlation > |0.7|. 27 species presented correlation > |0.7|; Table S1: List of ecoregions analyzed and original ecoregion names. Ecoregions with small geographic expressions have been joined to larger and closer ecoregions; Table S2: Justification for the inclusion of predictor variables used in species distribution models; Table S3: Parameters and hyperparameters used in each algorithm, values, and number of hyperparameter combinations, and R codes used to generate the sequence of values (.docx)

https://doi.org/10.21425/fob.18.142410.suppl1

Download file (1.78 MB)

﻿Abstract

﻿Highlights

Keywords

﻿Introduction

﻿Material and methods

﻿Study area

﻿Species list and occurrences

﻿Species distribution models and species richness maps

﻿Environmental variables

﻿Modeling protocol for species with > 15 occurrences

﻿Modeling protocol for species with 5–15 occurrences

﻿Modeling protocol for species with 2–4 occurrences

﻿Model validation and modeling postprocessing

﻿Species richness map

﻿Environmental heterogeneity variables and analyses

﻿Results

﻿Discussion

﻿Conclusion

﻿Acknowledgements

﻿Author contributions

﻿Data accessibility statement

﻿References