A global review of determinants of native bee assemblages in urbanised landscapes

Loss of natural habitat through land‐use change threatens bees. Urbanisation is a major, increasing form, of habitat loss, and a novel, pervasive form of disturbance known to impact bee diversity and abundance in a variety of often inconsistent ways. We conducted a comprehensive, semi‐quantitative review, involving 215 studies, on responses of bees to urban landscapes, and local and landscape variables proposed to influence bee abundance and diversity. Urban areas tend to be favourable habitat for bees compared with agricultural ones, but compared with natural areas, urban areas often host more abundant populations yet fewer species. Factors associated with urban landscapes, including changes in foraging resources and nesting substrate types and availability, contribute to changes in abundance, species richness, and composition of native bee assemblages. However, the conclusions of studies vary greatly because of the difference in the ecological traits of bees, habitats surveyed, and geographic region, as well as noise in the data resulting from inconsistencies in sampling methodology, and definitions of ‘urban’ and ‘natural’. Identifying what biotic and abiotic features of cityscapes promote or threaten the persistence of urban bee diversity is critical. We provide a comprehensive evaluation of how bees (both in aggregate and according to their ecological guild) have responded to the urban environment, identify gaps in knowledge in urban bee ecology, and make recommendations to advance our understanding of bees in urban environments to promote conservation of diverse bee communities.


INTRODUCTION
Bees are the most important pollinating group globally (Willmer et al., 2017) and the pollination services they perform are essential for stable, functioning ecosystems, both natural and anthropogenic (Potts et al., 2016). Although the domesticated European honey bee Apis mellifera L. is the most familiar and widely managed pollinator, wild bees, with an estimated global diversity of over 20,000 species (Ascher & Pickering, 2020), are critical to healthy ecosystems and are an essential component of biodiversity (Garibaldi et al., 2013;Potts et al., 2016).
There have been documented declines of bees across Europe, America, and China, particularly over the last 50 years (Bartomeus et al., 2013;Biesmeijer et al., 2006;Williams et al., 2009). But while there are concerns that bee declines are a global phenomenon , the conservation status of most bees is unknown (Nieto et al., 2014;Potts et al., 2016). The cause(s) of these declines include habitat loss, fragmentation and degradation, poor nutrition, disease, toxins, climate change, inappropriate disturbance regimes, and exotic species (Brown & Paxton, 2009;Goulson & Nicholls, 2016). Urbanisation can involve all of these threatening processes.
Urbanisation is considered a leading form of ecologically destructive global change (Elmqvist et al., 2016). Urbanised environmentslandscapes of human settlement that are created specifically for human occupation (Adler & Tanner, 2013;McIntyre et al., 2008) are the most heavily modified and rapidly expanding forms of anthropogenic land-use modification (Seto et al., 2011). By 2030, global urban expansion is predicted to increase by 285% over 2000 levels (Seto et al., 2012) and is occurring in regions known to harbour rare, endemic fauna (e.g. Ives et al., 2016;Phillips et al., 2010).
Urbanisation is now a major driver of fragmentation and loss of natural habitat (Winfree et al., 2007), and a key cause of biodiversity loss worldwide (Brown & Paxton, 2009), with bees predicted to be especially susceptible (Winfree et al., 2011). However, depending on patch quality and connectivity, and the surrounding matrix, urban areas have the potential to support a high diversity and abundance of native bees (Hinners et al., 2012) (Supplementary Materials S1).
This may especially be the case when comparing urban with agricultural assemblages, which contrasts highly complex landscapes typical of urban areas, with low-complexity landscapes typical of agricultural monocultures. Although urban and agricultural lands are both anthropogenic habitats, they differ abiotically and biotically (Table 1), reflected in distinctive bee assemblages (e.g. De Palma et al., 2015, 2016Sattler et al., 2011). Generalisations from bee responses to agricultural landscape modification should not, therefore, be extrapolated to urban landscapes.
Urbanisation can be predicted to influence native bee populations by altering the amount, quality, diversity, and distribution in space and time of the two most important resources for bees: flowers and nest sites. Urbanisation also alters the plant community composition, whereby native vegetation is often replaced by exotic plants. Urbanisation can be considered as a selective force filtering out species that are maladapted to urban conditions and selecting for 'synanthropic' species, which may reach high abundances (McKinney, 2006). Traits such as resource specialisation, body size, sociality, nesting substrate, and kleptoparasitism can be expected to determine the success of bee taxa in urban landscapes (Table 2) and influence species richness, abundance, evenness, community composition and functional and phylogenetic diversity of bee communities. These in turn can have implications for pollination; for example, functional trait diversity has been shown to maximise pollination (e.g. Fründ et al., 2013;Woodcock et al., 2019).
Evidence-based recommendations for bee-friendly management in cities are rare (notable exceptions include Bee City USA ® ; Bee City USA, 2021;Bee City Canada;Bee City Canada, 2021; and the European Commission's 'A Guide for pollinator-friendly cities ';Wilk et al., 2019). Moreover, planting guides are often focussed on the domesticated, often introduced and non-threatened European honey bee (e.g. Edmansun, 2021), giving honey bees a competitive edge and favouring this introduced species to the detriment of native bees . There are, however, promising opportunities to harmonise bee conservation with activities that promote ecosystem services and human welfare in cities. For improved management plans to be developed, there is a need to understand the current state of the science and where the knowledge gaps remain.
Here, we review publications on the responses of wild bees to urbanisation throughout the world. Although there have been reviews on, or that include, bees in urban areas (Cane et al., 2005;De Palma et al., 2015;Hall et al., 2017;Hernandez et al., 2009;Wenzel et al., 2019;Winfree et al., 2009;Winfree et al., 2011;Wojcik, 2009;Wojcik & Buchmann, 2012), ours is the most comprehensive to date with 215 studies reviewed in total. Moreover, our review is the first to conduct semi-quantitative analyses, investigating how abundance and species richness vary according to landscape type, local and landscape variables, and how responses vary among bee taxa with different functional traits. We also identify key knowledge gaps in understanding the determinants of urban bee abundance and diversity.

METHODS
From August 2016 to December 2019 searches were performed in Google Scholar using combinations of the terms: 'bees, pollinators, insects, arthropods, native bees, wild bees' combined with 'urbanisation, cities, urban, land-use change, suburban, metropolis'.
Google Scholar was chosen as it has less barriers than many other search engines (Martín-Martín, 2018: it is not restricted to users affiliated with a research institution and does not suffer from constrained coverage. Thus, our search included theses, books, reports, conference proceedings and articles, which may not be in mainstream English science journals (e.g. especially those published in developing countries) (e.g. Haddaway & Bayliss, 2015;Meneghini & Packer, 2007). This approach allowed us to achieve our objective of providing a comprehensive review on bees in urban areas.
Each text was incorporated if it included a measure of bee(s) abundance, reproduction or species richness in an urbanised landscape or response to urbanisation (however defined by the authors of the papersee results below and Supplementary Materials S2).
Papers included in this review therefore were either conducted in urban landscapes or in non-urban habitats that considered the influence of surrounding urban land use on bees.
(Supplementary Materials S1). Landscape and habitat types were assigned according to categories provided by original authors. For single-species publications, we excluded those on honey bees, as their numbers are largely impacted by husbandry (Champetier et al., 2015).
We could not attempt a formal meta-analysis due to the extreme variability in survey duration and number of sites surveyed, area sur- Nicholson & Egan, 2020).
We extracted the following information from each publication: the main findings; the type of study (whether it was of an entire community, a subset of the bee community, or focused on just one or a few species); details about the study design (geographic region, city, climatic zone); the number of sites surveyed; duration of the study (number of months per year, number of years), the sampling area, the sampling intensity, and the sampling method used; the landscape type (natural/rural/urban), and the urban habitat type(s) surveyed (see Figure 2b) (Supplementary Materials S1). For studies providing data on bees in urban landscapes as well as those in agricultural and/or natural landscapes, we recorded whether bee abundance and/or species richness was significantly different between these landscapes as T A B L E 2 Predictions on how bee functional traits will influence how bees respond (at a relative advantage or disadvantage) to urbanisation determined by the authors of these studies. We also extracted information on environmental variables that might influence bee communities and how these related to bee abundance and/or diversity. These related to both food and nesting resources, and landscape composition (Supplementary Information S1 and S5). We coded each variable in each study in terms of it having a positive, negative, or nonsignificant impact on bees (Supplementary Material S1). Some potentially important variables (pesticides, mowing and grazing, 'wildlifefriendly' gardening, human activity, landscape diversity, socio-economics, housing density, human density, road traffic) were investigated in too few studies to be included but are summarised in Supplementary Information S5.
We also tabulated abundance and species richness responses (positive, negative, non-significant) by the following ecological traits of bees: nesting substrate, kleptoparasite/host, sociality, body-size, lecty, origin (native/exotic), or higher-level bee taxonomic categories (family or genus). (Supplementary Materials S2 and S3). We note that there is a continuum of sociality but for ease of analysis, we categorised species as: solitaryone female per nest or social: any form of communal nesting or sociality with a reproductive division of labour (Wcislo & Fewell, 2017). We also investigated the community composition of bees in urban areas. We extracted information from each study in terms of the number of individuals recorded, the number of species, and number of genera. We also averaged the proportion of bee individuals and species in the different ecological trait categories listed above across urban bee studies. Many studies (47%) did not define explicitly what constituted an 'urban area'. For those that did, definitions varied widely (Supplementary Materials S2). How intensity of urbanisation was determined also F I G U R E 2 Breakdown of urban bee studies in terms of (a) survey methods; (b) urban habitat types surveyed. Note that some studies included more than one topic/habitat type or did not mention the habitat type other than just classifying it as 'urban'; hence, total numbers may not be the same as the total number of studies. Categories of habitat types in which native bee surveys and studies have been conducted (

Bees in urban versus natural and agricultural landscapes
There were 51 studies that compared bee communities between urban and rural/agricultural and/or 'natural' landscapes (as defined by the authors). The rest compared a variety of urban habitat types (Supporting Information S1, S5 and Figure S5). For urban and agricultural landscape comparisons, urban bee abundance was higher in 38% of cases, lower in 19%, and did not differ in 43% ( Figure 5). Likewise, for comparisons of urban and natural landscapes, urban bee abundance was higher in 32%, lower in 22%, but similar in 47% ( Figure 5).
For bee species richness comparisons between landscape types, urban areas had more species than agricultural landscapes (44% of cases) or did not differ (33%), with 22% of cases having fewer ( Figure 5). In contrast, species richness in natural areas was higher than in urban sites in almost half of the comparisons (48%), while urban areas had more species in 26%, and in 26% of cases, there was no significant difference ( Figure 5).  (Figure 6a,b). Approximately half of the studies, however, found no effect of flower abundance and richness on bee abundance (flower abundance: 45%, flower richness: 45%) or bee species richness (flower abundance: 56%, flower richness: 46%) (Figure 6a,b).

Determinants of native bee diversity and abundance in urban areas
When authors considered how native versus introduced plant species influence bees in terms of proportion of flowers or flower species that were native, native flora were generally beneficial, with 50% of cases finding positive associations with bee abundance and richness ( Figure 6), and most of the remaining associations were nonsignificant (bee abundance: 42%, bee species richness: 46%). There were therefore almost no negative associations (bee abundance: 8%, bee species richness: 5%).
Our data indicate that less groundcover (i.e. greater amounts of open, natural substrate) tends to be associated with increased bee abundance in urban habitats (38% of cases, compared with 6% of negative associations), and especially bee species richness (100% of cases) ( Figure 6). The proportion of area covered by trees had mainly positive (36%) or neutral (57%) associations with bee abundance (Figure 6a). Greater openness (i.e. the reciprocal of canopy cover, and an indicator of how much solar radiation a site receives) also tends to promote bee abundance (positive associations in 35% of cases vs. 12% negative associations) (Figure 6a). Increase in the area sampled ( Figure 6) had inconsistent effects on both bee abundance (36% positive, 22% negative and 44% non-significant associations) ( Figure 6a) and species richness (30% positive, 13% negative, 57% non-significant associations) (Figure 6b).
Although built space at landscape scales was more often negatively than positively correlated with bee abundance (negative associations in 40% of cases vs. 13% of positive associations) and richness F I G U R E 5 Studies involving comparisons between urban and natural habitat types that report no differences, a positive effect, or a negative effect on bee abundance and bee species richness for bees in urban compared with natural and agricultural/rural habitats. Number of studies assessed for each comparison above the columns (negative associations in 27% of cases vs. 17% of positive associations), in approximately 50% of cases, there was no significant impact (47% for abundance and 56% for species richness) ( Figure 6). Proportion of the urban area that was greenspace had more positive than negative impacts on bee abundance (31% positive and 12% negative,) and species richness (40% positive and 24% negative). Isolation from natural areas was often negatively and never positively associated with bee abundance (37% negative vs. 0% positive) and especially species richness (56% negative vs. 0% positive). Distance from city centres tended to have a positive association with bee abundance (67% positive, 33% neutral, 0% negative) and species richness (57% positive, 29% neutral); however, a small proportion of negative associations (14%) were recorded for species richness (Supplementary Materials S5 and Figure S5.
There were few strong patterns found for particular bee guilds and environmental variables, conceivably due to small sample sizes.
The raw data are tabulated in Supplementary Materials S3 and S4).
With more studies providing data on Apis mellifera and Bombus (either F I G U R E 6 Response of bees (positive, non-significant or negative) to major local and landscape level factors measured in urban bee studies in terms of (a) abundance and (b) species richness. Numbers above each column indicate the total number of studies involving that factor. Key: Flower N = flower abundance or density; Flower R = the species richness or diversity of flowering plant species; Native flora = the amount or proportion of native flowering plant species; Ground cover: the amount of ground cover, such as mulch, or grass at a site, or the reciprocal of open or bare ground; Openness: how open a site is, or the amount of solar radiation received, the reciprocal of canopy cover; Trees: the number of trees or woody plants, or tall plant forms; Area: the area or size of a habitat/patch; [the above are considered to be site scale variables, whereas the following are landscape scale: Built-space: a proxy of urbanisation, the proportion of built-space (impervious surfaces such as buildings, roads, pavement etc.)] around a site, measured at varying degrees of resolution and radii; Greenspace: proportion of vegetated area of any description, measured at varying degrees of resolution and radii; Isolation: how fragmented a site is or isolated from natural areas; note there is much variation in the number of studies that have investigated different explanatory variables. See Supplementary Materials S1 for full suite of variables focussed on these species or as a category within bee assemblage studies), clear favourable associations between their abundance and flower abundance were evident: for Apis, 10 positive, one negative and two non-significant associations and for Bombus 11 positive, one negative and five non-significant associations. These data can be contrasted with a predominance of neutral associations between flower abundance and that of all other (i.e. non-Apis and non-Bombus) bee taxa combined: six non-significant and four positive associations.
Therefore, it appears that general positive associations between flower abundance and bee abundance are more apparent for Apis and Bombus, whereas neutral associations feature to a greater extent for other bee taxa.

Ecological traits of urban bee assemblages
Oligolectic bees averaged 17% of species (

Bees in urban versus natural and agricultural landscapes
A key finding of our review was the importance of assessing abundance and species richness separately, as these metrics exhibited different patterns. While abundance was often higher in urban than natural landscapes, urban landscapes had fewer species. This suggests that a subset of species are benefitting in urban areas, while others are lost when natural areas are replaced with urban landscapes. However, urban areas tended to be better for bees than agricultural landscapes. This may be attributed to higher pesticide and herbicide use, homogenisation of the landscape, lack of suitable resources and floral monocultures (often of cereal-dominated wind-pollinated crops) in most agricultural contexts (Calatayud-Vernich et al., 2018;Goulson et al., 2015;Roulston & Goodell, 2011), which contrasts with the highly heterogeneous landscape and diversity of flora that can be found in some urban areas (Table 1).
Even though wild bees can be adversely affected by some facets of the urban environment (Table 1), some features of cities can be beneficial (Figure 7). For example, we found that most studies found positive rather than negative associations between bees and openness, and both flower abundance and richness. Thus, in comparison with dense, floristically poor wind-pollinated deciduous forest, flowerrich urban habitats (e.g. roadsides, waste-lands, and urban gardens) can be favourable for bees (Hall et al., 2017;Sirohi et al., 2015). By comparison, in natural habitats that have a diverse bee-friendly flora, native bees may respond less favourably to replacement of natural vegetation by urbanised landscapes (Martins et al., 2013).

Determinants of native bee diversity and abundance in urban areas and their composition
Although flower numbers had more often positive than negative associations with bees, most studies failed to find flower abundance, species richness, or even the proportion of flora that is native, to be associated with significantly greater abundance and/or species richness of bees in urban areas ( Figure 6 and Results above). It is also notable that over 13% of cases found a significant negative association between flower species richness and bee species richness ( Figure 6b). The lack of positive relationships may be because bees are more strongly correlated with the abundance of a few particularly attractive plant species than with plant diversity per se (Haaland et al., 2011;Lazaro & Totland, 2010;Prendergast et al., 2022;Rundlöf et al., 2014). Even generalists may benefit from single-species pollen diets if these have high protein, sterol and essential amino acid content (Di Pasquale et al., 2016;Moerman et al., 2017). Oligolectic bees are inevitably dependent on the limited plant taxa they forage on (Praz et al., 2008), so with native flora being replaced by exotic flora in urban areas, oligoleges may be expected to be less resilient to urbanisation. Lower effective population sizes of oligoleges also renders them at higher risk of extinction from stochastic and genetic events, reduced adaptability, and increased metapopulation extinction (Frankham et al., 2002;Zayed, 2009). Nevertheless, there is evidence that some oligolectic species can persist or even benefit under urbanisation, depending on how urbanisation alters their host plant availability (Cane et al., 2006). For example, a survey in Brazil assessing urbanisation impacts at a single site over 40 years recorded a small increase in the proportion of oligolectic species, even as bee richness and abundance declined overall (Martins et al., 2013).
Social bees require food over a greater proportion of the year than do solitary bees (Ogilvie & Forrest, 2017), so it is not surprising that abundances of social honey bees and Bombus, but not abundance of solitary bee taxa, were positively associated with floral abundance and diversity.
Bees of different nesting guilds might be expected to show different responses to urbanisation, due to how urbanisation impacts substrates for ground-nesting versus cavity-nesting taxa (Table 2, Supplementary Materials S3, S4) (Krombein, 1967). Although some studies have found that species nesting in small cavities are more abundant in urban than nearby natural areas (Cane et al., 2005), ground-nesting bees are generally more abundant and species rich than cavity-nesting species in urban areas (Supplementary Materials S1), although slightly less than the estimated global level of 83% Body size has also been predicted to influence the ability of a species to persist in urban habitats, but predictions vary regarding directionality of the response (Table 2). We found larger bodied bees were favoured relative to small-bodied bees both in proportion of individuals and species. Presumably, the ability to traverse the urban matrix to find rewarding patches favours larger species with greater foraging ranges (Merckx et al., 2018;Theodorou et al., 2021).
Predictions about how sociality influences bees in relation to urbanisation also vary ( Table 2). We found the average proportion of solitary species in urban areas was lower than the 75% of species globally (Danforth et al., 2019). However, it should be noted that authors varied in their categorisation of sociality especially in halictids, which exhibit variation in sociality, even within a species (e.g. Yanega, 1997). For example, Arena and Sgolastra (2014) considered Halictidae to be solitary, whereas most were classified as eusocial by Fetridge et al. (2008).
Kleptoparasite species appear to be relatively underrepresented in urban areas, comprising approximately 10% of species, less than worldwide estimates of 15% according to Wcislo (1996), and 13% according to Danforth et al. (2019). However, the abundance of kleptoparasitic individuals was similar to their relative abundances in natural habitats (e.g. Minckley, 2008). Kleptoparasites are predicted to be vulnerable to urbanisation, with complex consequences on hosts (Table 2). We found, however, that urban environments can still represent supportive habitats for this guild as has been found by others (Tscharntke et al., 1998;Sheffield et al., 2013).

KNOWLEDGE GAPS AND FUTURE RESEARCH DIRECTIONS FOR BEES IN THE URBAN CONTEXT
We identify seven broad knowledge gaps that should be fruitful areas for future research: F I G U R E 7 Aspects of the urban environment that can be beneficial for bees. Refer to Supplementary Materials S6 for references 3. Investigating how historical land-use influences the current composition of the bee fauna: Historical land use and the age of a city are likely to have a major influence on the composition of contemporary urban bee communities (Cusser et al., 2015), although there is also evidence that bees respond rapidly to landscape alterations both negatively and positively Onuferko et al., 2018). Older cities may have a bee fauna adapted to the city environment, whereas in recently urbanised areas, bees may have had insufficient time to adapt, while sensitive species may not yet have been eliminated due to lag effects (Ramalho et al., 2014;Ramalho & Hobbs, 2012). Such historical or legacy effects have not been sufficiently investigated when evaluating bee responses to urbanisation. What makes this a particularly attractive line of research is the detailed maps that are often available for cities for much of their history. Although ongoing monitoring using systematic methods is optimal (Marlin & LaBerge, 2001;Prendergast & Hogendoorn, 2021), in the absence of repeated surveys, museum collections represent a means of assessing changes in bee composition with ongoing urbanisation over time (e.g. Vaudo et al., 2018). Extinction debt under urban expansion has been considered for other pollinators (Soga & Koike, 2013), and future work on bees requires addressing the predictions that ongoing losses may occur in recently urbanised landscapes, whereas in landscapes with long histories of anthropogenic modification, we may now detect little effects of urbanisation.

Investing in long-term ongoing monitoring of bees in urban areas:
Studies involving long-term (>5 years) bee community changes under urbanisation have no clear consistent findings (Archer, 2013;Frankie et al., 2009;Martins et al., 2013) (Supplementary Materials S1). Moreover, these suffer from being comparisons in two points in time; ongoing systematic monitoring over time is required (Cane, 2001;Lebuhn et al., 2013;Packer & Darla-West, 2021). The limited duration of surveys also raises concerns over the reliability of conclusions given high variability in bee populations [Roubik, 2001;Williams et al., 2001;see Cane & Tepedino, 2001 for pitfalls associated with documenting declines over time and potential solutions for them]. As cities expand, how to best develop cities in harmony with biodiversity is a challenge.
There is some evidence that to optimise ecosystem services, land sparing (leaving aside large areas of land to nature i.e. in protected areas) is superior to land sharing (integrating 'wildlife friendly' practices into coupled human-nature landscapes) in urban ecosystems (Stott et al., 2015), and recent research indicates urban native vegetation remnants are crucial for conserving native bees (Prendergast & Ollerton, 2021a). That species richness is naturally high in areas undergoing rapid urban expansion underscores the importance of identifying how best to manage urban growth to preserve local bee biodiversity (Luck, 2007), and monitor bees under future urbanisation using an adaptive management approach.

Investigating how definitions of urban and the scale of analysis
influences results: Urban landscapes can differ markedly in impervious and 'greenspace' cover, design and configuration (Fuller & Gaston, 2009 (Glaum et al., 2017), 'village' habitats considered urban in the study by Samuelson et al. (2018) would be considered to fall outside this category, having on average less than 14% impervious cover.
Equally, the 'urban' sites in Nakamura and Kudo (2019) Bosch et al., 2000;Boyle & Pitts-Singer, 2017). The research to date on bee hotels in urban areas is limited but has found that they can be highly used by some species, but not others (Prendergast et al., 2020), and that caution needs to be exercised as in some locations they are highly used by introduced species (MacIvor & Packer, 2015). Apart from reducing sealed surfaces and artificial turf, identifying how to protect and create nesting substrates for urban ground-nesting bees should also be investigated (Fortel et al., 2016;Prendergast, 2021;Shepherd & Ross, 2003). 7. A better understanding of bee metapopulation structure maintenance in urban landscapes: Given the limited flight ranges of many bees (Zurbuchen et al., 2010), it cannot be assumed that the 'plant it and they will come' philosophy will apply (Tasker et al., 2019); proximity to healthy source populations or translocation may be the only practical means to reinstate native bees in restored/ designer landscapes. Because many bees are philopatric (Cane et al., 2006;Morato & Martins, 2006), it may be necessary to translocate native bees (Seddon, 2010), for the benefit of both bees and plants. Introducing species, especially outside their historical ranges, is a controversial idea (Kerr et al., 2015), that has been gaining traction under re-wilding initiatives and in the context of climate change. However, it is not without risks and must be planned very carefully (Lozier et al., 2015;Prendergast, 2020aPrendergast, , 2020b. Population genetic analyses are needed to understand how bees use the landscape for foraging and nesting, both at individual and metapopulation levels (e.g. Rosa et al., 2016). In particular, we need to know whether green corridors (such as power lines, road and railway verges) provide effective connectivity between patches in urban landscapes.
Because standard practices of weed removal, pesticide application, and mowing in most managed gardens reduce bee abundance and diversity (Rundlöf et al., 2015;Wastian et al., 2016), it can be predicted that wild bees within urban regions may depend on retention of 'unmanaged' natural areas in the urban matrix (Batista Matos et al., 2013;McFrederick & LeBuhn, 2006). Native vegetation in urban landscapes is the overriding factor in determining the persistence of unique and rare urban bee species (Geslin et al., 2016;Letourneau et al., 2012;Venturini et al., 2016). As source habitats with spill-over effects, they are a source of pollinators in more modified greenspace sinks (Hunter, 2002;Öckinger & Smith, 2007). Establishing or maintaining ecological connectivity between native remnant source populations will be important for enabling native bees to move throughout urbanised landscapes (Banaszak-Cibicka et al., 2016).
'Hotspots' where native bees are particularly abundant can be used to guide restoration of other sites, or even seed sites with pollinators; such sites must be prioritised for protection against further urban development (Prendergast & Ollerton, 2021b). However, identifying practices that maximise the value of anthropogenic greenspaces is also important because native vegetation remnants may be of insufficient size and connectivity to promote sustainable bee populations (Allen-Wardell et al., 1998; Rosenzweig, 2003).

Conserving wild bees in the urban jungle
From our review of the literature on bees in urban areas, the following recommendations can be made on how to improve urban areas to support native bee assemblages: • Retain, restore, revegetate and reconnect patches of remnant natural habitat throughout the urban matrix.
• Encourage homeowners, gardeners, landscape managers, and nurseries to focus on flowers that have been demonstrated to be visited by wild bees in the region. Flowers should be primarily native species and ensure oligolectic bee preferences are represented.
Key exotic flowers that have low risk of becoming invasive and offer quality and large volumes of nectar and pollen, especially outside the main flowering period of native flora, can also be included.
• Retain entomophilous street trees, as well as maintain open habitat.
• Leave patches of bare ground in greenspaces, and avoid plastic turf and large-scale mulching of gardens • Leave flowers and weeds on lawns, or even better, opt for flowering groundcover as a replacement for lawns.