Functional ecology of wild bees in cities: towards a better understanding of trait-urbanization relationships

In recent years there has been an increasing number of studies dealing with the functional diversity and trait–urbanization relationships of wild bees in urban environments. Indeed, many of the studies reviewed occurred in the last five to ten years. Most of the studies included ‘nesting type’, ‘diet’, ‘body size’, ‘sociality’ and ‘phenology’ as traits, with body size being the most common trait examined. More than a third of all of the studies we reviewed only described bee traits. These studies mostly calculated observed trait frequencies and described how frequencies changed across an urbanization gradient (e.g., with % urban land cover) or in different urban habitats (e.g. gardens vs. cemeteries), without testing for significant positive or negative relationships with respect to increasing urbanization or differences between environmental factors (Fetridge et al. 2008; Kearns and Oliveras 2009; Sirohi et al. 2015; Lerman and Milam 2016; Cardoso and Goncalves 2018). Regression analyses remain the dominant paradigm in urban wild bee ecological research. These analyses revealed that nesting behavior correlates with garden size (Quistberg et al. 2016: number of cavity species increase with garden size) or urbanization (Geslin et al. 2016a: decrease of ground-nesting bee abundance and species richness), and functional dispersion correlates positively with urbanization (Martins et al. 2017).

There are very few studies addressing functional diversity of wild bees that explicitly apply statistics on functional diversity indices and trait-environment relationships in urban areas, which is surprising given the widely recognised importance of functional analyses when relating biodiversity dynamics to environmental variables (Violle et al. 2007; Loreau 2010; Magurran and McGill 2011). We found only five studies that integrated explicit measures of functional diversity. Martins et al. (2017) analyzed the effects of urbanization and related habitat quality on ‘functional dispersion’ (the relative abundance of functional traits within a community, cf. Laliberté and Legendre 2010) using linear regression (e.g. Fig. 1b). They found that residential gardens can act as refugia from pesticides and offset the inadequate floral and nesting resources of agricultural landscapes. Semi-natural areas can thus support high levels of diversity and a bee fauna distinct from that of gardens. Normandin et al. (2017) applied Rao’s average quadratic functional diversity and their analysis demonstrated that urban cemeteries were less functionally diverse than urban community gardens and urban parks. Buchholz et al. (2020) assessed functional dispersion of wild bee assemblages in grasslands along an urbanization gradient in Berlin but found no direct effect of urbanization. Braaker et al. (2017) found an increase in Rao’s average quadratic functional diversity with increasing area and connectivity of urban green roofs, while Hung et al. (2019) showed that functional dispersion decreases in urban habitat fragments relative to natural habitats. Additional functional diversity metrics, such as functional evenness and functional divergence could be assessed with respect to urbanization because both consider other aspects of functional diversity (Schirmel et al. 2012, 2016). For example, urban habitats inhabited by functionally less even species communities may have a decreased productivity and reliability in terms of functioning and are more susceptible to invaders (Mason et al. 2005). Also communities with high functional divergence may have increased ecosystem function as a result of more efficient resource use (Mason et al. 2005). However, both have not been considered in urban wild bee studies to date, although both indices can be easily calculated using open-access statistical software (e.g., the R package FD, which can also calculate functional dispersion; Laliberté et al. 2014).

Few studies are going beyond traditional regression methods to analyze bee trait-urbanization relationships. Only two studies (Braaker et al. 2017; Harrison et al. 2018) applied the RLQ- and fourth-corner analysis to statistically test trait–environment relationships (e.g. Fig. 1c–e). This method is providing novel insights by making clear which traits are favoured in urban environments, which in turn determines which species become a successful urban dweller and which do not. It can be performed easily in the ade4 R statistical package (Dray and Dufour 2007; Dray et al. 2007) and creates comprehensive outputs including trait-environmental variable biplots that display the direction and statistical significance of trait–environment relationships (e.g., Buchholz et al. 2018, 2020; Fig. 1).

Several of the reviewed studies found lower abundance of ground-nesters in more “urban” habitats in comparison to more “rural” habitats including parks and gardens (Neame et al. 2013; Threlfall et al. 2015; Geslin et al. 2016a; Sivakoff et al. 2018; Twerd and Banaszak-Cibicka 2019). “Urban” habitats can be defined by, for example higher proportion of impervious surface or residential housing in the surrounding area than rural habitats (Kotze et al. 2011). Large areas of impervious surfaces and little semi-natural habitat in the landscape and compacted or covered soils might be detrimental to ground-nesting species (McIntyre and Hostelter 2001; McFrederick and LeBuhn 2006; Threlfall et al. 2015; Quistberg et al. 2016). These species instead respond positively to increased availability of forests and grasslands in their surroundings at the landscape scale (Pardee and Philpott 2014), and to soils with fine substrates (Kratschmer et al. 2018) and more bare soil available (Quistberg et al. 2016) at the local scale. Also abandoned rodent holes can provide valuable nesting structures for soil nesting bees (e.g. bumble bees) (McFrederick and LeBuhn 2006). As such, urban environments such as urban parks or gardens where rodent activity is prevalent can provide habitat for ground nesters that provide pollination services. Above-ground cavity-nesting wild bees can be more frequent in urban compared with rural habitats (Cane et al. 2006; Matteson et al. 2008; Fortel et al. 2014). Cane et al. (2006) stated that cavity-nesting wild bees are less affected by urbanization than other nesting guilds because urban areas provide numerous suitable nest resources such as fencing, homes, walls and shade trees close to the floral resources of gardens (Pardee and Philpott 2014). However, Threlfall et al. (2015) argued that the nesting success of cavity nesters depends upon the cavity substrate and the ability of species to utilize certain man-made materials. Together, the relationships between nesting traits and environmental factors suggest that habitat management at the local scale is an important filter of bees with certain traits related to nesting behaviour. Specifically, managing ground surface substrate and diversity as well as tree or tree-like cavities at the local and landscape scale is important for supporting bee species of more diverse nesting behaviours across urbanization gradients.

There is no strong consensus on the relationship between urbanization and wild bee diet-related traits, including oligolecty (i.e. specialized feeders) and polylecty (i.e. generalist feeders). Two studies found no differences between urban and natural habitats regarding wild bee diet (Gotlieb et al. 2011; Wray et al. 2014). Although oligolectic wild bees may be abundant in some urban habitats such as gardens or parks (Frankie et al. 2005; Banaszak-Cibicka et al. 2018a, b; Hamblin et al. 2018; Lerman and Milam 2016), most studies found an overall reduction in the abundance and species richness of oligolectic wild bees in cities in comparison to rural or natural/semi-natural areas (Cane 2005; Cane et al. 2006; Fetridge et al. 2008; Banaszak-Cibicka and Zmihorski 2012; Zurbuchen and Müller 2012; Kratschmer et al. 2018, Twerd and Banaszak-Cibicka 2019) while polylectic species are more abundant in cities in comparison with rural landscapes (Matteson et al. 2008; Ahrné et al. 2009; Antonini et al. 2013; Hausmann et al. 2016; Jedrzejewska-Szmek and Zych 2013; Deguines et al. 2016). These studies propose that this relationship is likely due to the abundance of ornamental plant species in cities that attract pollinators. Collectively these results suggest that plant–pollinator networks seem to be less diverse in cities because they comprise few specialists and many generalists (Martins et al. 2017).

There is some evidence across the literature that urbanization filters for small-bodied bees. Some studies showed that wild bees are smaller in urban compared with rural habitats (Ahrné et al. 2009; Banaszak-Cibicka and Zmihorski 2012; Wray et al. 2014; Hamblin et al. 2018; Eggenberger et al. 2019), and large-bodied species are often scant in cities (e.g., [Metro-Detroit Region, MI, USA] Glaum et al. 2017; [Poznan, Poland] Banaszak-Cibicka et al. 2018a, b).

Eggenberger et al. (2019) analyzed urban bumblebee species and stated that smaller body sizes in cities can be related to effects of resource availability and heat islands effects on species physiology. Resource availability is often limited in urban sites with high amounts of impervious surfaces in the surroundings (e.g., Colla and MacIvor 2016), and during species larval stages undernutrition can lead to smaller body sizes in adults (Sutcliffe and Plowright 1988; Schmid-Hempel and Schmidt-Hempel 1998). Furthermore, smaller body sizes in urban wild bee populations may be related to urban heat.

Negative relationships between bee species body size and increasing urbanization found in these studies may be related to bees responding to habitat and resource availability at small spatial scales. Body size of wild bees is also linked to foraging range, in which large-bodied species are able to fly longer distances in search of resources while smaller bees have smaller foraging ranges (Araújo et al. 2004; Stang et al. 2006; Greenleaf et al. 2007). Small-bodied bee species might need less pollen and nectar to successfully reproduce (Cane et al. 2006). Also, there may be no need to fly long distances to forage in cities because of the very diverse and abundant floral resources in close proximity to one another often planted in ornamental landscapes including home gardens, community gardens, and city parks (Lowenstein et al. 2014; Hülsmann et al. 2015; Hall et al. 2017).

There is no generalizable trend in sociality across urban habitats. Some studies found social species are more frequent in urban habitats surrounded by a high amount of impervious cover compared to suburban or rural sites (Harrison et al. 2018; Banaszak-Cibicka et al. 2018a; Kratschmer et al. 2018). Several studies found no significant urbanization effects on the abundance of eusocial species (Zanette et al. 2005; Carper et al. 2014; Guenat et al. 2019). Urban areas may provide more diverse nesting substrate for social colony building species that can utilize features of the built environment (Cane et al. 2006). In addition, the presence of honey bees in urban areas may also negatively impact the foraging and reproductive capacity of solitary species (e.g. Ropars et al. 2019), thus outcompeting solitary species. However, given the limited data further work is needed to test these hypotheses around sociality and urbanization.

Although there are no generalizable relationships between phenology and seasonality and urbanization, some studies suggest that wild bees with later spring emergence dates and later flight seasons seem to be better-adapted to urban environments. For example, Harrison et al. (2018) observed increased abundance of wild bees with late flight seasons in urban habitats relative to forest habitats, due to the presence of late-blooming plants. Banaszak-Cibicka and Zmihorski (2012) found that wild bees with later emergence dates prevail in urban settings while spring-emerging bees may be negatively affected by urbanization due to urban warming if this changes flowering phenology (Wray et al. 2014; Harrison and Winfree 2015). Spring-emerging wild bees that overwinter as adults might lose more weight before emergence than summer-emerging bees overwintering as larva, likely because of their higher metabolic activity in response to warmth (Fründ et al. 2013). Some pollinators may adapt and lengthen their flight season in urban habitats, as found for Bombus terrestris, which collected nectar and pollen during a mild winter (Stelzer et al. 2010) and also Harrison et al. (2019) found longer flight seasons for wild bees in urban habitats. Floral resources provided by non-native species flowering at different times than natives or blooming for long periods may extend activity periods of wild bees (Harrison and Winfree 2015).

Trait-based ecology of wild bees in urban environments has received much attention in recent years, but we do not have strong evidence of generalizable trait shifts with increasing urbanization. In this section, we discuss potential reasons for inconsistencies and multi-directional responses between bee traits and urbanization across the current literature. Specifically, we focus on three research methodological biases that limit our understanding: (1) bias in wild bee species traits; (2) bias in sampling design; and (3) bias in statistical approach.

Across the current literature there is a bias in the environmental data, specifically around the metrics of urbanization. Our analysis is hinged upon the relationship between urbanization at the landscape scale and wild bee traits, yet the “urbanization” variable is not consistently measured across studies. In several studies, the quantification of urbanization was expressed as either (a) proportion of impervious surface in a buffer around the study site (Buchholz et al. 2020; Eggenberger et al. 2019), (b) residential land use (Lerman and Milam 2016), or (c) proximity to urban development (Kearns and Oliveras 2009). A lack of a definition of urbanization or measurement of urbanization is a drawback because environmental effects are not comparable across different studies which hinders generality and synthesis of functional relationships. Furthermore, only a few studies covered the whole gradient from urban to rural sites with a quantification of urbanization (mostly proportion of impervious surface in a buffer around the study site) (e.g., Buchholz et al. 2020, Eggenberger et al. 2019, Guenat et al. 2019). The majority of studies compared sites only along a narrow gradient (e.g., Carper et al. 2014, Martins et al. 2017, Twerd and Banaszak-Cibicka 2019), sites within one urbanization category (e.g., Pardee and Philpott 2014, Threllfall et al. 2015, Quistberg et al. 2016), or without properly defining urbanization (e.g., Geslin et al. 2016b, Sivakoff et al. 2018, Harrison et al. 2019). In addition are the possible interactions overlooked between urbanization land cover factors and other abiotic (e.g. urban heat) (Hamblin et al. 2017) or biotic (competition with honey bees) (Ropars et al. 2019) drivers and also habitat quality can modulate relationships (Buchholz et al. 2020).

Inconsistencies across different studies may arise either from a geographical bias or a habitat bias of the wild bee research to produce a study system bias. First, most studies are conducted in the Global North. This suggests not only a geographical bias, but also a climate regime bias, which is relevant because temperature is an important driver of trait shifts (Hamblin et al. 2017). Also, a geographical bias may result from several studies from one research group that may bias the literature towards that very specific system or geographic region.

Furthermore, our review suggests a habitat bias since most studies have been conducted in gardens, which often have high ornamental richness and abundance and therefore are different in vegetation composition from urban wastelands, lawn or cemeteries (Francis and Chadwick 2013)—all habitats that were underrepresented in our review. Ideally there is a need for equal representation to draw conclusions.

Sampling method can strongly affect results in trait research because sampling methods bias detection towards particular species and away from others (Wong et al. 2019). Some researchers used window traps, bee bowls, direct observation, hand sampling, or photographic surveys. Sweep netting may produce biases because observers may not be the same, or because sweep netting is not possible at all sites simultaneously. Observational and/or photographic surveys can produce biases because it is very difficult to generate species-level identification needed for functional trait analyses. Although pan trapping may have biases in which species it attracts (Roulston et al. 2007) this method may reduce observer bias and any temporal bias, and can produce data at the trait level.

In ecological research, there is often a trade-off between replication (number of study sites) and sampling intensity (number of times visited, duration of trap exposition) (e.g., Coddington et al. 1991; Oksanen 2001). Our review suggests that it is often challenging to find a sufficient number of suitable sites in cities to sample for wild bees because replication per urbanization category or habitat is often below five and in many cases unbalanced. To maximize replication, we recommend to, on the one hand, focus on a standardized type of model ecosystem (e.g., von der Lippe et al. 2020) and, on the other hand, particularly compare urban and rural sites and to go without intermediate urbanization levels. If pan trap methods are applied, such a focused study design can ensure high replication with an appropriate sampling intensity.

We conclude by proposing three main knowledge gaps: (1) extending urban geographical and study system breadth; (2) increasing consistency and transparency across trait-based research and accounting for interspecific trait variation; and (3) linking urban wild bee traits to ecosystem function.

First, there is an urgent need for studies from Africa, Asia, and the Mediterranean region of Europe. South America and Australia are also underrepresented, considering these areas contain biodiversity hotspots. In addition, most studies have focused on wild bee traits within urban gardens, but there are many more valuable habitat types in cities that should be investigated. In particular, vacant lots and wastelands with spontaneous vegetation, amenity grasslands and even roadside greenery should be considered as these habitat types contribute considerably to the urban green infrastructure and have been shown to harbour functionally diverse wild bee communities.

Second, future studies should increase the consistency of urban wild bee trait-based data collection, analysis and transparency. Here, investigators should collect primary data on traits in a continuous form because categorical traits are less suitable to reflect changes in trait-environment interaction (Lavorel and Garnier 2002). This approach can best capture intraspecific trait variation present in most terrestrial invertebrates but largely underexplored (Didham et al. 2016). Furthermore, all of this trait data should be open-source for all researchers to inform their investigations and results (see Wong’s (2019) recommendations for the development of terrestrial arthropod databases). Once trait data are collected, studies should apply appropriate trait-based statistic approaches to explicitly analyze the relationships between wild bee functional traits and urban environmental variables, rather than simply present observational data. Regression, or even better, RLQ- and fourth-corner analyses are very useful and widely accepted statistical tools in ecology to directly statistically correlate environmental variables to trait frequencies by combining trait, species, and environmental data matrices (Dray et al. 2014). This may better inform the currently missing generalizable conclusions regarding trait shifts with environmental variables in urban areas. Indeed, our review’s sample size was limited to half of all studies that met our initial search criteria due to the lack of rigorous statistical methods.

Third, future research should relate urban wild bee traits to ecosystem function and link traits to pollination of wild plant species along with cultural plants prevalent in cities (e.g. ornamental and food crops). Linking these traits to a function, and subsequently likely to an ‘ecosystem service’, can also better guide or heighten people’s awareness about the role of wild bee (functional) diversity in their cities.