Urbanization decreases richness but not the phylogenetic and functional diversity of bees

doi:10.21203/rs.3.rs-4173630/v1

Download PDF

Article

Urbanization decreases richness but not the phylogenetic and functional diversity of bees

https://doi.org/10.21203/rs.3.rs-4173630/v1

This work is licensed under a CC BY 4.0 License

Version 1

posted

You are reading this latest preprint version

Urbanization has been documented as a major factor affecting worldwide biodiversity. Such evidence, however, mostly came from studies of plants and birds, with the response of many taxa to urbanization unclear. For example, both increases and decreases in bee diversity with urbanization have been reported. Such studies mainly focused on a small number of cities, with no study attempting to assess the patterns of bee diversity with urbanization at large spatial scales. To fill this knowledge gap, we leveraged global bee occurrence data to quantify species richness, phylogenetic diversity, and functional diversity of bees across Canada and the United States. We analyzed the effects of urbanization, climate, and agriculture on bee diversity while controlling for the number of records, ecoregion, and spatial correlations. We found that bee richness decreased with urbanization. In addition, urbanization led to the loss of redundant phylogenetic lineages, increasing the phylogenetic distances among closely related species. Urbanization also increased the occurrence of species with uncommon functional traits, increasing functional diversity. Future studies should expand these analyses to more regions, including the speciose areas across Mediterranean climates, as well as Tropical regions.

Biological sciences/Ecology/Urban ecology

Biological sciences/Zoology/Entomology

Biological sciences/Ecology/Biodiversity

Biological sciences/Ecology/Biogeography

Hymenoptera

Apoidea

Global Change

MPD

MNTD

Faith’s PD

species traits

species lineages

North America

city

Urbanization is one of the principal manifestations of the Anthropocene and a key driver of global environmental change ¹. Urbanization caused 16% of the natural habitat loss between 1992 and 2000 ²; since then, it has been expanding and intensifying, housing 55% of the human population in 2018 ^2–4. Urban growth is not only impacting newly urbanized areas, but also surrounding ecosystems ⁵, driving extensive changes in natural communities and global biodiversity through alterations in the natural cycles of water and temperature ⁶, pollution ⁷, spread of exotic species ⁸, and degradation, destruction, and fragmentation of natural soils and habitats ⁹.

The effects of urbanization on biodiversity are complex and context-dependent, varying across taxonomic groups and geographical context ⁸. On the one hand, urbanization impacts natural patterns and processes, causing local extirpations of many species. On the other hand, urbanization benefits some local species by providing copious and novel resources, extending natural activity periods, and releasing species from some of their natural enemies ¹⁰. Hence, urbanization studies are prone to report context-dependent or contradictory results for several taxonomic groups, including bees (Hymenoptera: Apoidea).

Bees are essential taxa for ecosystem functioning because they are pollinators for more than two-thirds of all extant plants ^3,11. Bees are also critical urban species since they pollinate plants in ornamental and farming gardens ³. Despite their relevance for global and urban ecosystems, there is no consensus on the impacts of urbanization on bee communities. While some studies have reported positive or negative responses in bee abundance and richness, these responses tend to be context-dependent and can vary based on factors like local habitat characteristics and regional landscape matrices ^8,11. Additionally, many of these studies have focused on a limited number of cities, which can further complicate our understanding of the broader impacts of urbanization on bees ^8,11.

Despite the context-dependent effects on abundance and richness, the few studies examining bee phylogenetic and functional diversity have consistently reported declines due to urbanization ^3,12. Studies examining the effects of urbanization on phylogenetic diversity have reported losses of phylogenetic lineages and overall community homogenization ^3,12. These trends suggest that urbanization may have significant impacts on the evolutionary history and ecological function of bee communities. Urbanization is also associated with the targeted selection of generalist and social bees that nest above-ground, given their capacity to exploit multiple floral resources (including exotic plant species), their greater adaptability and persistence over time, and the detriment of below-ground nesters due to the spread of non-impervious surfaces and urban soil management ^3,8,13. The effects of urbanization on bee size are less clear, with studies reporting size increases and decreases ^8,13,14. However, similar to the studies on bee richness and abundance, published studies have either studied the effects of urbanization for a single city, or used multiple studies from single cities to assess overall trends. No studies have assessed the effects of urbanization on bee richness, phylogenetic or functional diversity at large geographical scales.

Large-scale assessments can evaluate how urbanization affects bee species richness among cities and test the generality of phylogenetic and functional diversity losses. Historically, studies at large geographical scales have been constrained by data availability, but bee occurrence data is becoming rapidly available, especially in Western countries such as the U.S.A. and Canada. A recently developed workflow used bee data from global, regional, and local datasets to obtain a clean list of world bee occurrences (BeeBDC) ¹⁵. We leveraged the BeeBDC dataset to study the effects of urbanization on bee biodiversity within Canada and the continental U.S.A., dividing the study area into 10 km by 10 km grid cells and evaluating grid cell bee species richness, phylogenetic and functional diversity. We studied functional diversity based on the most commonly used bee functional traits, related to bee morphology, diet, and habits. We then evaluated the relationship between bee diversity and urbanization (based on human population density and built surface) as well as other important ecological drivers, including climate (temperature, precipitation, and solar radiation), and land use (cropland areas and landscape heterogeneity) while controlling for the effects of ecoregion, spatial correlations, and sampling intensity (number of observations).

Our analysis included 2,553 species of bees from all North American bee families (Andrenidae, Apidae, Colletidae, Halictidae, Megachilidae, and Melittidae). Bee phylogenetic and functional distances were moderately correlated (Mantel test: r = 0.27, P-value = 0.010) (Fig. 1), and functional traits displayed significant phylogenetic signals (assuming evolution based on Brownian motion) (female body length: Pagel’s λ = 0.892, P-value < 0.001; male body length: Pagel’s λ = 0.888, P-value < 0.001; lecty: D-value = 0.144, P-value (random) < 0.001, P-value (Brownian) = 0.154; sociality: D-value = -0.849, P-value (random) < 0.001, P-value (Brownian) = 0.999, kleptoparasitism: D-value = -0.948, P-value (random) < 0.001, P-value (Brownian) > 0.999).

Taxonomic diversity

Bee richness decreased with urbanization, precipitation, and land use diversity, and increased with solar radiation, proportion of croplands in the landscape, and sampling intensity (Fig. 2, Table S.1.1). Bee richness was also significantly related to temperature, peaking at mild temperatures. Land use, climate, and sampling intensity explained 76.3% of the variance in bee richness, while ecoregion and the geographical association among samples explained an additional 2.1%. Therefore, bee communities included more species in less urbanized areas with sunny, mild, and arid climates, as well as closer to farmlands and within more homogeneous landscapes. The results were equivalent when limiting the analysis to native species (Supplementary Material Appendix S.3).

Phylogenetic diversity

The indices of bee phylogenetic diversity responded differently to urbanization (Fig. 4, Table S.1.2). Faith’s phylogenetic distance (Faith’s PD), which quantifies phylogenetic lineage richness and is intrinsically related to species richness, decreased with urbanization. Mean phylogenetic distance (MPD), which assesses the average distance among lineages and is independent of richness, did not vary significantly with changes in urbanization. Mean nearest taxon distance (MNTD), which measures the number of unique lineages and compactness of phylogenetic lineages, increased with urbanization. Therefore, urbanization leads to the loss of bee phylogenetic lineages, but affects lineages that are mostly redundant within the community, maintaining average phylogenetic distances among species and even increasing the minimum pairwise distances among taxa. A 10-fold increase in urbanization (product of population density and proportion of built land) led to the loss of almost 100 million years of phylogenetic lineages and an increase of 4 million years of phylogenetic distance among closely related species.

Climate had consistent effects on bee phylogenetic diversity; phylogenetic diversity peaked at mild to warm temperatures, lower precipitation, and higher solar radiation. However, the effect of precipitation was not significant for MNTD. MNTD increased with landscape heterogeneity. Additionally, MNTD decreased with the proportion of cropland, whereas Faith’s PD increased. Samples with more records had higher overall phylogenetic diversity, but lower MNTD. Urbanization, climate, land use, and sampling intensity explained about 35% of the variances of MPD (33.3%), and MNTD (40.3%), and around 70% of the variance of Faith’s PF (66.1% raw, 68.8% of log-transformed Faith’s PD). Ecoregion and geographical distance explained an additional 20.5%, 3.2%, 3.6%, and 6.8% of the variances of MPD, MNTD, and the raw and log-transformed Faith’s PD. The models for standardized effect sizes show that urban areas include more lineages than expected by the species loss, as well as negative effects of sampling intensity, a significant effect of precipitation on SES MNTD and significant effects of the proportion of cropland (negative) and landscape heterogeneity (positive) on Faith’s PD (Supplementary Material Appendix S.4). The models including only native species show qualitatively similar results, without the relationship between population density and MNTD (Table S.3.2).

Bee families had distinct responses to urbanization, climate, and land use (Fig. 5, S.1.1). Apidae bees tended to be present in densely populated and highly urbanized areas, whereas species of Andrenidae and Melittidae were more common in rural areas (Fig. 5). Bee families also differed on their environmental preferences for other climate and land use variables (Fig. S.1.1).

Functional diversity

Urbanization slightly increased the functional diversity of bee communities. A 10-fold increase in urbanization (i.e., population density x proportion of built land) led to approximately 1% increase in the functional distance between species and closely related taxa (mean MPD = 26%, mean MNTD = 7%). Additionally, bee functional diversity increased with precipitation, and land use diversity, and decreased at intermediate temperatures.

MPD increased with solar radiation and sampling intensity, whereas MNTD decreased (Fig. 6). Additionally, MNTD decreased with the proportion of surrounding croplands. Urbanization, climate, land use, and sampling intensity together explained, respectively, 13.0, 38.3 and 42.5% of the variances of MPD and raw and square root-transformed MNTD, whereas ecoregion and geographical locations explained an additional 9.1, 2.3 and 2.4%. Hence, bee traits are most diverse in densely populated and highly urbanized areas, high precipitation, and diverse surrounding landscapes and less diverse in areas with mild temperatures. The standardized effect sizes (SES) for the models reveal quite consistent trends, except for a negative effect of the proportion of croplands on SES MPD, and positive effects of solar radiation and sampling intensity for SES MNTD (Supplementary Material Appendix S.4). The models excluding non-native species were qualitatively similar except for the effects of temperature on MPD, and solar radiation on the square root-transformed MNTD (Supplementary Material Appendix S.3).

Urban areas were favored by either opportunistic or above-ground nesters, polylectic, eusocial, and free-living bee species. Urbanization also favored larger bee species, displaying similar effects on both female and male lengths (Fig. 5). Bee functional traits also mediated species responses to other climatic and land use variables (Figs. S.1.2 – S.1.6). For the most part, precipitation and landscape heterogeneity had similar effects on bee functional traits as urbanization, favoring social, polylectic, and larger species, whereas temperature, solar radiation, and proportion of cropland had opposite effects. Social, polylectic, large bee species were more common in urbanized areas with high precipitation, and diverse surrounding landscapes, as well as cold and less sunny areas with few croplands.

We found a decline in bee richness with urbanization, even when excluding exotic bee species. However, the loss of bee richness in urban areas was mostly due to the disappearance of redundant phylogenetic lineages, rather than the loss of phylogenetically distinct species. In fact, more urbanized areas showed slightly higher bee functional diversity, which could be attributed to the occurrence of uncommon bee functional traits, such as above-ground and opportunistic nesters, polylectic, and social species. The lack of speciose clades of urban-associated and urban-deterred species may explain the effect of urbanization on phylogenetic diversity, despite the consistent loss of bee richness.

The decline of bee richness in urban areas is consistent with most studies on bee communities and urbanization ^8,11,16. The decrease of bee richness in highly urbanized areas can be explained by the lower availability of floral and nesting resources, high fragmentation and isolation of suitable patches, and spread of non-native plant species ^8,11,16. Bee richness also increased in warmer, arid, and sunny areas, consistent with prior studies ^17,18. Globally, bee richness peaks in warm and semi-xeric climates with high solar exposition ^17,18. These conditions promote productive systems with many co-flowering insect-pollinated species (e.g., prairies and meadows), and may increase competition and specialization of bee species ¹⁸. Additionally, lower precipitation allows for bee species to store larval food with lower risk of fungal infection and decreases the need for anti-flooding measures ^18,19.

Contrary to prior studies, we found that the proportion of croplands in the landscape increased bee richness ^8,11. Indices of taxonomic diversity that include species abundances, like those used in other studies, may be more suited to reveal declines in bee diversity in croplands; however, our bee occurrence data are mostly unstructured and incidental, limiting our ability to infer unbiased abundances from the observations. Additionally, most studies reporting bee richness declines in croplands compared heavily agricultural and urban sites, whereas our grid cells included few croplands (median = 2% of the cell, mean = 16%, std. dev. = 25%). Small farms may increase floral resource diversity and availability, and provide diverse nesting substrates. On the other hand, we found a negative effect of land use heterogeneity on bee richness, which may be associated with a greater presence of less suitable habitats (e.g., closed canopy forests, or water bodies).

Urbanization has a complex effect on bees' phylogenetic diversity. While urban areas have consistently lost lineages along with bee richness, the overall average phylogenetic distances among species have remained intact, with redundant rather than rare lineages being lost (lower Faith’s PD, invariant MPD, and higher MNTD). The analyses of standardized effect sizes of phylogenetic diversity revealed consistently higher phylogenetic diversity than expected by the species losses. Similar trends of Faith’s PD, MPD, and MNTD have been reported using abundance-based diversity metrics ^{20 but see 12}, so that urbanization may especially deter redundant species and individuals. The effects of climate on bee phylogenetic diversity were, overall, consistent with the effects on species richness. Similar overlaps were found between bee species richness and endemism ¹⁷. Our results indicate that areas with mild to warm, semi-dry, and sunny climates may not only contain more bee species, and endemisms, but also phylogenetic lineages. We did not find consistent effects of cropland proportion on bee phylogenetic diversity, in contrast to a prior study ²¹. However, these differences may be explained by differences in the geographic scope of the studies, with the prior study being based in fewer, heavily managed areas ²¹.

Bee communities in urban areas included more diverse functional traits. Bees in urban areas were more likely to display less common functional traits, such as nesting above-ground or indistinctly below and above-ground, feeding on many plant species, and living in organized societies. The association between these uncommon functional traits and urbanization is supported by most prior studies ^8,11,22. Ground-nesting bees have limited nest availability in urban areas due to the spread of non-impervious surfaces, whereas cavity nesting bees can be favored by novel nesting strata (e.g., buildings, electric posts) ^8,11,13,22. Generalist bees are more common in urban areas possibly in response to changes in the local flora and/or higher competition for limited resources ^8,13. Social bees are common in urban areas either because they are more abundant, or because they are adaptable and can easily find resources ^{8,20 but see 13}. Additionally, we found kleptoparasites and smaller bees to be less common in urban areas. Parasitic bees tend to be less common in urban areas due to their dependence on appropriate hosts ^11,22. While the effect of urbanization on bee size is still unclear ^8,13,16,20, we found larger species to be more common in urban areas, possibly due to their ability to travel further and move within fragmented landscapes ^8,11.

The effects of climate on bee functional diversity differed from its effects on bee richness and phylogenetic diversity. Functional diversity increased with precipitation, but did not vary with solar radiation. Precipitation affiliations among species with different functional traits did not differ as dramatically as urbanization affiliations, but portrayed similar patterns. Humid areas deterred ground-nesters and favored polyleges and larger and social species, similar to urban areas. Despite possessing adaptations to reduce water infiltration, ground-nesters in wet climates may experience molding of larval resources, and lower larval survival ¹⁹. Additionally, areas with less precipitation may promote grasslands and pastures with more asynchronous flowering plants, foster competition, and promote oligolecty ²³. Fewer studies have assessed the effects of precipitation on sociality and bee size, with studies reporting both positive and negative effects of precipitation on solitary bees ²⁴ and no effects on body size ²³. Croplands reduced bee functional diversity, promoting common bee functional traits (i.e, ground-nesting, oligolectic, and solitary bees). Interestingly, the proportion of croplands was the only factor for which kleptoparasites had higher associations, suggesting that areas with a slightly higher proportion of croplands may enable the survival of very specialized bees.

Our study is, to the best of our knowledge, the first to report large-scale effects of urbanization on bee diversity. Additionally, our findings support many of the consistent trends revealed in the most recent and complete reviews. Further analyses should aim to incorporate species abundances, since urbanization is likely to impact evenness of bee species, lineages, and traits. Furthermore, future efforts should aim to increase the geographical coverage of the data, to limit geographical biases, and include urban areas with different histories and configurations as well as regional pools with different species richness. Two regions of interest include areas with Mediterranean and Tropical climates. Mediterranean climates favor bee species, and tend to include the most diverse assemblages of bee species ^17,18. While our study included the West Coast of the U.S.A., studies over other Mediterranean regions could add more clarity, including those in Northern Chile and Argentina, the Mediterranean Coast, Southern Africa, and Australia. These areas include diverse bee faunas and similar climates but cities with different histories, configurations, and current development. On the other hand, Tropical areas tend to include diverse bee lineages and functional traits, and could offer further insights related to the interactions between urbanization and latitude, and the effects of current urbanization on bee diversity.

Bee richness in North America North of Mexico decreases with urbanization. Lost bee species tend to have redundant phylogenetic histories in the community, so that the overall phylogenetic diversity is maintained across urbanization gradients, while the phylogenetic distance increases among closely related taxa. Functional diversity increases with urbanization, because cities deter species with common functional traits, including ground-nesters, oligoleges, and solitary bees. Future works should include additional regions and evaluate the effects of urbanization on abundance-based diversity indices. Understanding the effects of urbanization on bee diversity is essential to conserving bee assemblages and maintaining pollination services in an increasingly urbanized world.

Bee occurrence records

We collected publicly available records of bees (Hymenoptera: Apoidea) in Canada and continental U.S.A. from the BeeBDC database ¹⁵, 2,664,435 observations from 3,361 species. We then excluded spurious observations, such as those with unflagged geospatial problems (e.g., incomplete or unlikely coordinates), as well as paleontological records and records lacking basis information (i.e., whether the observation was based on a living or preserved specimen). We then excluded observations with invalid or problematic basis, coordinates, coordinates reprojection, datum, date, URI reference, or type status, retaining 2,540,129 records from 3,353 species. We selected all unique records by taxon ID, full name, family, species, references, and coordinates, resulting in 964,887 unique records. We cleared all taxonomic issues, using the taxonomic backbone in Discover Life ²⁵. We also excluded all bee records from Hawaii (i.e., 1,409 records of 44 species), as well as any observations of exotic species not registered in North America (12,432 records of 72 species). We excluded 274 bee records from 6 species whose coordinates had no associated urbanization or climate information (e.g., points over water). We selected species with at least 5 independent records (excluding 1,408 observations and 714 species), and grid cells that included at least 5 observations (excluding 83,109 observations). The final dataset included 866,247 unique records from 2,553 species. The data cleaning process was conducted in R v. 4.2.0 using packages ‘rgbif’ ²⁶, ‘scrubr’ ²⁷, and coordinateCleaner ²⁸.

Taxonomic diversity

We divided the study area in 10 x 10 km² grid cells, and assessed taxonomic diversity as the cell bee species richness. Species richness is just a simple count of present species that does not require abundance or sampling effort data, making it convenient for analyses using synthetic data such as BeeBDC, especially since it contains data from unstructured recreational observations without sampling effort data. We used species richness to estimate taxonomic diversity while controlling for sampling intensity (i.e., number of records) in our statistical models.

Phylogenetic diversity

To estimate phylogenetic diversity we used the maximum-likelihood tree of bees provided by Henríquez-Piskulich, Hugall, and Stuart-Fox (2023) ²⁹, one of the most recent and complete global bee phylogenies. We added missing species as polytomies at the basal node of their congeners, since many bee species lack extensive genomic data ³⁰. Additionally, we added four missing genera as polytomies of their sister genera: Melissoptila (tribe Eucerini ³¹), Microalictoides (related to Diadasia, Conanthalictus, Protoduforea, Sphecodosoma, and Xeralictus ³²), Neopasites (paraphyletic in respect to Biastes ³³) and Rhopalolemma (tribe Neolarrini ³⁴). The grafting of missing species was conducted with R package ‘rtrees’ v.1.0.1 ³⁵. Phylogenies derived in this way are reliable for calculating community phylogenetic diversity ³⁶.

We assessed three common metrics of phylogenetic diversity (PD), namely, mean pairwise distance (MPD ³⁷), mean nearest taxon distance (MNTD ³⁷), and Faith’s phylogenetic diversity (Faith’s PD ³⁸). Mean phylogenetic distance (MPD) refers to the mean branch length separating each pair of species in the sample. Mean nearest taxon distance (MNTD) is the average distance between each species in the sample and its closest neighboring species (i.e., the closest species in the tree present in the sample). Faith’s phylogenetic diversity (Faith’s PD) is defined as the sum of the branch lengths connecting the sample species across the phylogenetic tree. We calculated all PD metrics with R v. 4.2.0 and packages ‘ape’ v.5.6.2 ³⁹ and ‘PhyloMeasures’ v.2.1 ⁴⁰.

Functional diversity

We evaluated five functional traits to estimate functional diversity, including body length (separately for females and males), trophic specialization, nesting location, sociality, and kleptoparasitism (Table S.7.1) (Table 1). These are important traits that determine several functional aspects, including bee floral visits, habitat use and availability, phenology, exposition to flooding, and risks of overheating, parasitism, and predation ^{8,14,19,41,42}. These traits are also susceptible to urbanization, given the effects of urbanization on flower composition and distribution, nest availability, and prevalence of diseases and parasitism ⁸. We used Gower distances to compute the functional distance matrix since they can handle combinations of quantitative and qualitative traits, as well as missing data ^43,44.

Table 1

Description of the selected functional traits. We obtained the measurements for each functional trait from Moreno-García, Nguyen, and Li (2024).
Functional traits	Measurement	Rationale
Female length (mm)	Average or mid-point (for ranges) of female body length (head to last tergite)	Related to flower accessibility, resistance to overheating, predation ¹⁴. Bee size is altered by urbanization; however, the direction of this effect is unclear ^8,14.
Male length (mm)	Equivalent to female body length	We distinguished between male and female lengths due to bee sexual dimorphism ⁶⁵.
Trophic specialization	- Oligolectic/ Specialized - Polylectic/ Generalized	Describes bee diet breadth or lecty ⁴². Trophic specialization of non-kleptoparasitic species is based on pollen sources, that of kleptoparasites is based on nectar sources). Urbanization tends to benefit polylectic species ⁸.
Nest location	- Below-ground - Above-ground - Both (hereafter ‘Opportunistic’)	Determines habitat availability as well as the species exposition to flooding, predators, and other risks ¹⁹. Urbanization tends to benefit above-ground nesting species ⁸.
Sociality	- Eusocial - Solitary or primitively social (hereafter ‘Solitary’)	Influences resource gathering, distribution area, annual activity cycle, and predation risk ⁴¹. Urbanization tends to benefit social species ⁸.
Kleptoparasitism	- Kleptoparasitic - Non-kleptoparasitic	Parasitic bees lay their eggs in the nests of other species to avoid parental care ¹⁹, having reduced nesting and resource gathering costs ⁶⁶. Parasitic bees tend to have a narrow host selection which makes them less common in anthropogenic habitats ²².
¹Intertegular distance (ITD) is probably a better estimator of size since it is clearly defined and not highly dependent on the position of the pinned specimen. However, ITD information is especially scarce for species first described long ago and not included in very recent taxonomic keys.

We used two metrics to quantify functional diversity (FD): mean pairwise distance (MPD ³⁷), and mean nearest taxon distance (MNTD ³⁷) based on functional traits. We calculated the functional MPD and MNTD using the functional distance matrix as well as a species presence matrix. Functional MPD and MNTD are equivalent to phylogenetic MPD and MNTD but rely on functional distances rather than branch lengths. We calculated all FD metrics with R v. 4.2.0 and packages ‘cluster’ v.2.1.3 ⁴⁵, and ‘picante’ v.1.8.2 ⁴⁶.

Independent variables

We studied the response of bee diversity to urbanization, which was measured using human population density and built surface ⁴⁷. However, our models would be spurious if we ignored the effects of other important environmental drivers such as climate ¹⁸ and agriculture ^21,48. Hence, we studied the effects of urbanization, climate, and land use on bee diversity (i.e., bee richness, phylogenetic, and functional diversity). We assessed the climatic conditions by quantifying the effects of temperature, precipitation, and solar radiation ¹⁸. To quantify land use conditions, we calculated cropland proportion as well as the diversity of land uses within each grid cell. Additionally, we controlled the effects of other confounding variables, including the number of records, and the spatial correlation, and ecoregion of the samples.

We measured urbanization as the product between population density and percent of built surface. We obtained human population density data from the Global Human Settlement Layer for 2020 at a resolution of 1 km (GHSL ⁴⁹). We selected the data for 2020 since most of our records were recorded during that period. We re-projected the points at a resolution of 10 km to match the bee occurrence data and extracted the values of human population density for each grid cell. We obtained the built surface, cropland proportion, and land use diversity from Dynamic World ⁵⁰ using the average map from 2020, which classified each grid cell into one of nine categorical land uses (i.e., water, trees, grass, flooded vegetation, crops, shrub and scrub, built, bare, and snow and ice ⁵¹). We used cropland proportion and land use heterogeneity since both agriculture and habitat heterogeneity have extensive effects on bee diversity ^11,14,48. For the proportion of built area and cropland, we counted the 0.01 x 0.01 km² (10 x 10 m²) Dynamic World grid cells contained within each 10 x 10 km² grid cells that were classified as cropland or built surface using Google Earth Engine ⁵¹. For the diversity of land uses, we used the Shannon-Weaver index ⁵² to calculate the diversity of land uses among the 0.01 x 0.01 km² grid cells within each of the 10 x 10 km² grid cells using Google Earth Engine. We obtained the climate data from WorldClim v.2.1 ⁵³. We downloaded the climate data for 1970–2000 at a resolution of 2.5’ (approximately 5 km at the equator). While the temporal period covered by WorldClim is not ideal for the study, it contains geographically consistent measurements of average temperature, precipitation, and solar radiation. We reprojected the data at a resolution of 10 km and extracted the climate data at the grid cell center. The centers of some grid cells were located on water bodies, so we used the average over the coordinates of the bee occurrences within those grid cells instead. We obtained the map of ecoregions level II from the US EPA portal ⁵⁴. We extracted the ecoregion value for the center of each 10 x 10 km² grid cell. Some grid cells did not return any values (e.g., located within water bodies). Thus, we used buffers of increasing size to assign ecoregion values and visually resolved all the cases in which a grid cell was added to either multiple or no regions.

Statistical methods

To evaluate the effects of urbanization, climate, agriculture, and landscape heterogeneity on bee diversity, we used linear mixed models with the bee diversity metrics as response variables and land use, climate, and sampling intensity explanatory variables. To account for spatial autocorrelations among grid cells, we included Gaussian correlations among the coordinates of grid cell centers in the model and included a random effect of ecoregion at the grid center. We fitted six models, one for each of the bee diversity metrics (species richness, phylogenetic MPD, MNTD, and Faith’s PD, and functional MPD, and MNTD). Each model included seven explanatory variables: urbanization (product of population density and built surface proportion), temperature, precipitation, solar radiation, cropland proportion, Shannon-Weaver diversity of land uses, and number of bee occurrences (as an estimate of sampling intensity). We used variance inflation factors (VIF) to detect potential collinearity between covariates and found that all VIF were below 2 ⁵⁵ but those of solar radiation (3.5, correlation with temperature = 0.64, correlation with precipitation = -0.57) and temperature (2.7). We decided to maintain temperature and solar radiation in the model following a prior analysis of global bee diversity ¹⁸. We log-transformed urbanization, cropland proportion, and bee sampling intensity. We scaled and centered all explanatory variables (mean = 0, std. dev. = 1). We calculated the marginal and conditional R² using Nakagawa and Schielzeth’s method ^56,57. We repeated the analyses excluding exotic species to assess the robustness of our results to species introductions (introductions are more common in urban areas ³). We also repeated the models using standardized effect sizes (SES) of phylogenetic and functional diversity based on species pool within each ecoregion level II ⁵⁴ to control for the regional bee richness. The results for the models excluding exotic species and SES metrics, as well as the raw models with outliers are located respectively, in Supplementary Materials Appendices S.3-S.5. The models for the SES for functional diversity were very consistent, whereas those for phylogenetic diversity revealed some interesting discrepancies. Any important differences between the main results and those of the additional models are discussed in the results section.

To test whether different bee families and species with different functional traits are associated with different environmental conditions, we used linear mixed models with the values of the environmental variables as response variables, the categorical bee families and functional traits as explanatory variables (bee family, nesting location, trophic specialization, sociality, and kleptoparasitism), and ecoregion level II as a random intercept. We also evaluated the effect of the environmental variables on bee length, by using female and male lengths as response variables, the environmental variables as explanatory variables, and ecoregion as a random intercept.

We assessed the phylogenetic conservatism of bee functional traits to examine possible correlations between bee phylogenetic and functional diversity. To study phylogenetic conservatism, we evaluated the correlation between the phylogenetic and functional distance matrices, and computed the phylogenetic signal of all studied functional traits but nesting location (since the estimation of phylogenetic signals for categorical variables with more than 2 levels is non-trivial ⁵⁸). We evaluated the phylogenetic signal of female and male total length using Pagel’s λ, assuming Brownian motion ⁵⁹, and using package ‘phyloint’ v.0.1 ⁶⁰. We computed the phylogenetic signal of trophic specialization, sociality, and kleptoparasitism using the D index, assuming Brownian motion ⁶¹, and using package ‘caper’ v.1.0.2 ⁶².

For all models, we evaluated model assumptions, including normality of the residuals, homoscedasticity, and presence of outliers and influential observations. We performed all the analyses using R v.4.2.0 and packages ‘nlme’ v.3.1.157 ⁶³ and ‘MuMIn’ v.1.47.1 ⁶⁴.

Data availability

The database of bee occurrences that support the findings of this study are available in Dorey et al. (2023) figshare with the identifier: https://doi.org/10.25451/flinders.21709757.

The phylogenetic tree of bee (Hymenoptera : Apoidea) that supports the phylogenetic diversity analyses is available in Henríquez-Piskulich, Hugall, and Stuart-Fox (2023) Dryad Digital Repository with the identifier: https://doi-org /10.5061/dryad.80gb5mkw1.

The functional trait database that supports the functional diversity analyses is not yet publicly available, since it is currently in consideration for publication. The data is currently available as a supporting material not for publication, and will be available in a published article as well as a public database.

Code availability

The code for reproducing the analyses will be available upon article approval.

Acknowledgements

We want to thank Quynh Nguyen and Ariel Mata for their help gathering bee functional traits. This study was supported by the start-up fund to DL provided by Louisiana State University.

Contributions

P.M.G. conceived the idea, with input by D.L. P.M.G. performed the analysis. P.M.G. drafted the first version of the manuscript. L.C.P. and D.L. provided substantial input on the manuscript and analytical framework. P.M.G. and D.L. provided parts of the data. All co-authors edited the manuscript and provided suggestions on how to improve the analyses.

Competing interests

The authors declare no competing interests.

Kondratyeva A et al (2020) Urbanization Effects on Biodiversity Revealed by a Two-Scale Analysis of Species Functional Uniqueness vs. Redundancy. Front Ecol Evol 8:73
Ren Q et al (2023) Impacts of global urban expansion on natural habitats undermine the 2050 vision for biodiversity. Resour Conserv Recycl 190:106834
Ayers AC, Rehan SM (2021) Supporting Bees in Cities: How Bees Are Influenced by Local and Landscape Features. Insects 12:128
Theodorou P, Baltz LM, Paxton RJ, Soro A (2021) Urbanization is associated with shifts in bumblebee body size, with cascading effects on pollination. Evol Appl 14:53–68
Chen W, Chi G (2022) Urbanization and ecosystem services: The multi-scale spatial spillover effects and spatial variations. Land Use Policy 114:105964
Cuthbert MO, Rau GC, Ekström M, O’Carroll DM, Bates AJ (2022) Global climate-driven trade-offs between the water retention and cooling benefits of urban greening. Nat Commun 13:518
Brunekreef B, Holgate ST (2002) Air pollution and health. Lancet 360:1233–1242
Wenzel A, Grass I, Belavadi VV, Tscharntke T (2020) How urbanization is driving pollinator diversity and pollination – A systematic review. Biol Conserv 241:108321
Ferreira CSS, Walsh RPD, Ferreira AJ (2018) D. Degradation in urban areas. Current Opin Environ Science Health 5:19–25
Leong M, Ponisio LC, Kremen C, Thorp RW, Roderick GK (2016) Temporal dynamics influenced by global change: bee community phenology in urban, agricultural, and natural landscapes. Glob Change Biol 22:1046–1053
Prendergast KS, Dixon KW, Bateman PW (2022) A global review of determinants of native bee assemblages in urbanised landscapes. Insect Conserv Divers 15:385–405
Harrison T, Gibbs J, Winfree R (2018) Phylogenetic homogenization of bee communities across ecoregions. Global Ecol Biogeogr 27:1457–1466
Buchholz S, Gathof AK, Grossmann AJ, Kowarik I, Fischer LK (2020) Wild bees in urban grasslands: Urbanisation, functional diversity and species traits. Landsc Urban Plann 196:103731
Theodorou P, Baltz LM, Paxton RJ, Soro A (2020) Urbanization is associated with shifts in bumblebee body size, with cascading effects on pollination. Evol Appl 14:53–68
Dorey JB et al (2023) A globally synthesised and flagged bee occurrence dataset and cleaning workflow. Sci Data 10:747
Theodorou P (2022) The effects of urbanisation on ecological interactions. Curr Opin Insect Sci 52:100922
Bystriakova N, Griswold T, Ascher JS, Kuhlmann M (2018) Key environmental determinants of global and regional richness and endemism patterns for a wild bee subfamily. Biodivers Conserv 27:287–309
Orr MC et al (2021) Global Patterns and Drivers of Bee Distribution. Curr Biol 31:451–458e4
Antoine CM, Forrest JRK (2021) Nesting habitat of ground-nesting bees: a review. Ecol Entomol 46:143–159
Villalta I, Bouget C, Lopez-Vaamonde C, Baude M (2022) Phylogenetic, functional and taxonomic responses of wild bee communities along urbanisation gradients. Sci Total Environ 832:154926
De Palma A et al (2017) Dimensions of biodiversity loss: Spatial mismatch in land-use impacts on species, functional and phylogenetic diversity of European bees. Divers Distrib 23:1435–1446
Banaszak-Cibicka W, Dylewski Ł (2021) Species and functional diversity — A better understanding of the impact of urbanization on bee communities. Sci Total Environ 774:145729
Pardee GL et al (2022) Life-history traits predict responses of wild bees to climate variation. Proceedings of the Royal Society B 289, 20212697
Bottero I et al (2023) Impact of landscape configuration and composition on pollinator communities across different European biogeographic regions. Front Ecol Evol 11:1128228
Pickering J (2009) Discover Life. University of Georgia Athens
Chamberlain S et al (2022) ropensci/rgbif: rgbif v3.7.0. https://doi.org/10.5281/ZENODO.6023735. Zenodo
Chamberlain S (2016) scrubr: Clean biological occurrence records
Zizka A et al (2019) Standardized cleaning of occurrence records from biological collection databases. Methods Ecol Evol 10:744–751CoordinateCleaner
Henríquez-Piskulich P, Hugall AF, Stuart-Fox D (2023) A Supermatrix Phylogeny of the World’s Bees (Hymenoptera: Anthophila). http://biorxiv.org/lookup/doi/10.1101/2023.06.16.545281 doi:10.1101/2023.06.16.545281
Hedtke SM, Patiny S, Danforth B (2013) N. The bee tree of life: a supermatrix approach to apoid phylogeny and biogeography. BMC Evol Biol 13:138
Michener CD, LaBerge W, Moure J (1955) Some American Eucerini bees. Dusenia 6:213
Dumesh S, Sheffield CS (2012) Bees of the genus Dufourea Lepeletier (Hymenoptera: Halictidae: Rophitinae) of Canada. Can J Arthropod Identif 20:1–36
Bossert S et al (2020) Phylogenomic and Morphological Reevaluation of the Bee Tribes Biastini, Neolarrini, and Townsendiellini (Hymenoptera: Apidae) With Description of Three New Species of Schwarzia. Insect Syst Divers 4:1
Sless TJ et al (2022) Phylogenetic relationships and the evolution of host preferences in the largest clade of brood parasitic bees (Apidae: Nomadinae). Mol Phylogenet Evol 166:107326
Li D (2023) rtrees: an R package to assemble phylogenetic trees from megatrees. Ecography e06643 10.1111/ecog.06643
Li D et al (2019) For common community phylogenetic analyses, go ahead and use synthesis phylogenies. Ecology 100
Webb CO, Ackerly DD, McPeek MA, Donoghue MJ (2002) Phylogenies and Community Ecology. Annu Rev Ecol Syst 33:475–505
Faith DP (1992) Conservation evaluation and phylogenetic diversity. Biol Conserv 61:1–10
Paradis E, Claude J, Strimmer KAPE (2004) Analyses of Phylogenetics and Evolution in R language. Bioinformatics 20:289–290
Tsirogiannis C, Sandel B (2016) PhyloMeasures: a package for computing phylogenetic biodiversity measures and their statistical moments. Ecography 39:709–714
Ogilvie JE, Forrest JR (2017) Interactions between bee foraging and floral resource phenology shape bee populations and communities. Curr Opin Insect Sci 21:75–82
Gathmann A, Tscharntke T (2002) Foraging ranges of solitary bees. J Anim Ecol 71:757–764
Alb C (2021) Probabilistic Alternative to the Gower Distance: A Note on Deodata Predictors. 10.13140/RG.2.2.26146.35522
Gower JC (1971) A General Coefficient of Similarity and Some of Its Properties. Biometrics 27:857
Maechler M, Rousseeuw P, Struyf A, Hubert M, Hornik K (2022) cluster: Cluster Analysis Basics and Extensions
Kembel SW et al (2010) Picante: R tools for integrating phylogenies and ecology. Bioinformatics 26:1463–1464
Li D, Stucky BJ, Deck J, Baiser B, Guralnick RP (2019) The effect of urbanization on plant phenology depends on regional temperature. Nat Ecol Evol 3:1661–1667
Le Féon V et al (2010) Intensification of agriculture, landscape composition and wild bee communities: a large scale study in four European countries. Agriculture Ecosystems Environment 137:143–150
Pesaresi M et al (2016) The global human settlement layer from landsat imagery. in. IEEE International Geoscience and Remote Sensing Symposium (IGARSS) 7276–7279 (IEEE, Beijing, China, 2016). 10.1109/IGARSS.2016.7730897
Brown CF et al (2022) Dynamic World, Near real-time global 10 m land use land cover mapping. Sci Data 9:251
Gorelick N et al (2017) Google Earth Engine: Planetary-scale geospatial analysis for everyone. Remote Sens Environ 202:18–27
Shannon CE, Weaver W (1946) The Mathematical Theory of Communication. University of Illinois Press, Urbana
Fick SE, Hijmans RJ (2017) WorldClim 2: new 1-km spatial resolution climate surfaces for global land areas. Int J Climatol 37:4302–4315
Omernik JM, Griffith GE (2014) Ecoregions of the conterminous United States: evolution of a hierarchical spatial framework. Environ Manage 54:1249–1266
Zuur AF, Ieno EN, Elphick CS (2010) A protocol for data exploration to avoid common statistical problems. Methods Ecol Evol 1:3–14
Nakagawa S, Schielzeth H (2013) A general and simple method for obtaining R ² from generalized linear mixed-effects models. Methods Ecol Evol 4:133–142
Johnson PC (2014) Extension of Nakagawa & Schielzeth’s R2GLMM to random slopes models. Methods Ecol Evol 5:944–946
Borges R, Machado JP, Gomes C, Rocha AP, Antunes A (2019) Measuring phylogenetic signal between categorical traits and phylogenies. Bioinformatics 35:1862–1869
Pagel M (1999) Inferring the historical patterns of biological evolution. Nature 401:877–884
Eklöf A, Stouffer DB (2016) The phylogenetic component of food web structure and intervality. Theor Ecol 9:107–115
Fritz SA, Purvis A (2010) Selectivity in mammalian extinction risk and threat types: a new measure of phylogenetic signal strength in binary traits. Conserv Biol 24:1042–1051
Orme D et al (2013) The caper package: comparative analysis of phylogenetics and evolution in R. R package version 5:1–36
Pinheiro J et al (2017) Package ‘nlme’. Linear nonlinear mixed Eff models version 3:336
Barton K, Barton MK (2015) Package ‘mumin’. Version 1:439
Stubblefield JW, Seger J (1994) Sexual dimorphism in the Hymenoptera. differences between sexes 71–103
Litman JR, Praz CJ, Danforth BN, Griswold TL, Cardinal S (2013) Origins, evolution, and diversification of cleptoparasitic lineages in Long-tongued bees: evolution and diversification of Cuckoo bees. Evolution n/a-n/a 10.1111/evo.12161

There is NO Competing Interest.

BeeTraitsPublicationClean.csv
Table S.6.1 (Not for publication)
NCOMMS2414994rs.pdf
Reporting Summary
NatCommFSSupplementary.docx

Download PDF

Version 1

posted

You are reading this latest preprint version

Urbanization decreases richness but not the phylogenetic and functional diversity of bees

Status:

Version 1

Abstract

Figures

Introduction

Results

Discussion

METHODS

Declarations

References

Additional Declarations

Supplementary Files

Status:

Version 1