Temperature seasonality most strongly associated with Atlantic Forest bee richness

An ecological modeling study of native bees found that temperature seasonality was the strongest environmental variable associated with species richness across the Atlantic Forest, while topographic and geomorphic heterogeneity also showed positive biome-wide associations. The relationships varied among ecoregions, and environmental heterogeneity explained relatively little of the richness pattern in some of the most bee-rich areas.

The researchers modeled the distributions of 490 native bee species across the biome, which encompasses eastern Brazil and portions of eastern Paraguay and northeastern Argentina. They ultimately retained 466 species whose models met a Sorensen performance threshold of 0.7 or higher. Species with at least 15 occurrences were modeled using an ensemble of eight distribution algorithms, species with five to 14 occurrences were assessed with ensembles of small models, and distributions for species with two to four occurrences were estimated from environmental similarity within 50 km of occurrence records.

The occurrence data began with 240,062 records associated with 770 species names. After taxonomic corrections and selection of species native to the study area, the list included 564 species; 74 species represented by only one occurrence were removed before distribution modeling. Spatial cleaning removed invalid or duplicate coordinates, records georeferenced to administrative centroids and occurrences outside the documented natural distribution, reducing the dataset to 16,542 records. Environmental filtering to reduce sampling bias retained 11,404 occurrences.

Models were built from 1-km bioclimatic and elevation data that were resampled to 5 km. Principal components explaining 95% of the original environmental variation were used as predictors, with four components applied to 282 species, five to 206 species and six to two species. The richness map was created by stacking semi-binary models in which suitability below a threshold was assigned zero and values above the threshold remained continuous. Predictions were limited to a 100-km buffer around a minimum convex polygon enclosing each species’ presence records.

The resulting map placed the highest estimated richness, above 75 species per 5-km cell, in the southwestern and southern parts of the Atlantic Forest, including mountainous portions of the Serra do Mar Coastal Forests, eastern Araucaria Moist Forest and southern Upper Paraná Atlantic Forest. Estimated richness was lower in the northeastern coastal strip, where Pernambuco Coastal Forests contained 25 to 50 species per cell, and in drier interior areas of the Caatinga and Bahia Inland Forests, where values fell below 25 species. These are modeled richness estimates rather than direct counts from standardized field surveys.

Poisson generalized linear models related richness to four environmental measures: temperature seasonality, topographic heterogeneity, geomorphic heterogeneity and stream density. Spatial eigenvector filters were used to account for residual spatial autocorrelation, and roughly 2,200 cells were systematically sampled for each analysis. For the biome as a whole, temperature seasonality had a variable-importance value of 87.44, compared with 25.7 for topographic heterogeneity and 23.6 for geomorphic heterogeneity. All three were positively associated with modeled richness, while stream density had comparatively little explanatory power.

At the ecoregion scale, the strength and direction of these associations were not uniform. Temperature seasonality ranked as the most important variable in nine of the 11 ecoregions, including the Cerrado, Uruguayan Savanna and Campos Rupestres Montane Savanna. The researchers reported positive relationships in five ecoregions and negative relationships in six. Topographic heterogeneity was positively associated with richness in seven ecoregions and negatively associated in two, whereas geomorphic heterogeneity also changed direction among ecoregions. These nonstationary results indicate that a biome-wide relationship cannot be assumed to apply uniformly at the regional scale.

Notably, the Serra do Mar Coastal Forests had the highest estimated bee richness even though the environmental heterogeneity variables explained little of its richness pattern. The authors identified several possible considerations, including historical climatic stability, the quality and year-round availability of floral resources, and whether the 5-km modeling resolution was too coarse to capture finer environmental variation. These were proposed explanations rather than variables directly tested as causes of the regional pattern.

The study also examined a possible circularity because temperature seasonality was used in both the species distribution models and the community-level richness analysis. Only 27 species, or 5.8% of the modeled set, had suitability values strongly correlated with temperature seasonality, and they represented a small portion of total richness in the southwest. The authors therefore considered the potential influence of circularity to be minimal, although this check did not eliminate all limitations.

Important constraints include the difficulty of obtaining comparable occurrence records for bees across large and inaccessible areas, differences in species-specific life histories and habitat requirements, and the use of partly merged sections of the extensive Cerrado and Caatinga ecoregions. Consequently, the results should not be extrapolated to those ecoregions in full. Overall, the analysis supports temperature seasonality, terrain complexity and geomorphic diversity as factors associated with Atlantic Forest bee richness, while emphasizing that regional variation and factors not included in the models may also shape local diversity patterns.

Subject of Research: Environmental heterogeneity and modeled native bee richness across Atlantic Forest ecoregions

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

Article References: Antonio, A. I., Oliveira Junior, A. C. D., Villalobos, F., & Velazco, S. J. E. (2025). Environmental heterogeneity as a determinant of bee diversity patterns in the Atlantic Forest. Frontiers of Biogeography, 18, Article e142410. https://doi.org/10.21425/fob.18.142410

Image Credits: AI Generated

DOI: 10.21425/fob.18.142410

Keywords: Atlantic Forest, native bees, species richness, temperature seasonality, topographic heterogeneity, geomorphic diversity, species distribution models, ecoregions, spatial modeling, biodiversity, Environmental, heterogeneity