Data on the three interventions were collected in 2020 within the citizen-science SSNC campaign “Operation: Save the bees” using an online questionnaire prepared by the lead author (AP) in collaboration with officers at the SSNC. The campaign started in 2018 when volunteers in Sweden could register pollinator-friendly interventions that they had carried out in their gardens and other green spaces. The campaign encouraged three different interventions: (i) flowering “garden meadows,” (ii) bee-friendly flower plantings, and (iii) bee hotels. Those who registered that they had undertaken an intervention received an email with a link to questionnaires with queries regarding their intervention(s) at the end of the growing season in September 2020 (Table 1 and Supplementary material). Separate surveys were provided for each type of intervention. Hence, if a participant registered more than one type of intervention they received, and potentially answered, two or three separate surveys. The surveys were sent to all who had registered interventions between 2018 and 2020. Note that respondents could register interventions that had been established before the start of the campaign in 2018.

In the questionnaires, the participants were asked about flower-visiting insects in general, i.e., potential pollinators. Hereafter, we refer to them as pollinators. Participants were not asked to distinguish between different insect taxa. We used the questions related to the abundance of pollinators observed, the size, age, and floweriness or flowering plant species richness of the intervention, and the type of surrounding habitat for further analyses, Table 1. The assessment of abundance of insects in flower interventions was answered as either: “no insects,” “a few,” “many,” or “I do not know,” Table 1. Note that participants were not given any instruction on the definitions of “a few” and “many” insect pollinators.

Responses to the number of insects seen stating “I do not know” were removed from further analyses, as were interventions accidently stated to have been established before year 1900 or after September 2020, responses with an incorrect number of digits for year of establishment, and responses stating zero or >50 flowering plants species in flower interventions. One bee hotel, listed with a 1 million nest holes (a straw roof) was removed prior to analysis, resulting in a range of bee hotels with 1–2,500 nest holes. We also excluded responses where the number of occupied nest holes exceeded the total number of nest holes listed for the bee hotel, or where the number of nest holes per size category did not match the total number of nest holes listed. Collectively, this resulted in 370 responses being removed prior to analyses.

The number of insects seen in meadows and plantings was transformed into a binomial variable for further analyses, where 0 = No, or few, insects seen, and 1 = Many insects seen. We used year of establishment to infer age of intervention as a numerical factor 1–5, where interventions established before 2016 were merged into the oldest category (5) and years 2017–2020 were kept as four separate categories (4, 3, 2, 1). For the number of flowering plant species, responses of zero species were removed and the remaining responses were categorized in to five-step intervals: 1–5, 6–10, 11–15, 16–20, and >20 species. To assess potential effects of surrounding environment, the three non-garden categories (agricultural landscape, forestry landscape, and nature, Table 1) were merged into category rural. Gardens of urban single-family houses and allotment gardens were merged into category urban garden. Balconies and yards of multi-family houses were merged into category dense urban. Single-family rural gardens were kept as a single category, hereafter called rural garden. Thus, in total four environment categories were used for further analyses.

For the evaluation of how successful the respondents perceived their intervention to be (1–5, Likert scale), answers were grouped into three categories: low success (1–2), medium success (3), and highly successful (4–5).

To evaluate the accuracy of the simple assessment of pollinator abundance (none, few, or many insects seen in flower interventions), we used data from 2021 collected through another online questionnaire. Similar to 2020, this questionnaire was sent to all participants in the campaign 2018–2021, asking them to assess the abundance of insects in their intervention(s). In 2021, however, participants were also asked to complete a 10-min survey of 50 m² of their garden, which included their flower intervention, and to count all flower visiting insects into five groups (bees and wasps, hoverflies, butterflies, beetles, and other insects). The survey was to be performed sometime between 11.00 and 16.00 on a calm, sunny, and warm day (>16°C) in July.

We modeled to what extent flowering interventions (meadows and plantings) were visited by pollinators using generalized linear models (GLM), with binomial error distribution. We specified separate models for meadows and plantings. The proportion of participants who stated they observed “many insects” (as opposed to “none, or few, insects”), were modeled as a function of intervention area (categorical), intervention age (numeric: 1–5), species richness of flowers (categorical), and the type of surrounding environment (categorical). We assessed if flower intervention age and plant species richness was correlated using Spearman rank correlations, for meadows and plantings separately.

We modeled bee hotel occupancy using a GLM, specified with a beta binomial distribution. Occupancy was modeled as a function of environment (categorical), nest size category (categorical: 2–5, 6–10, and 11–15 mm wide), degree of flowering (numeric: 1–5), and bee hotel age (numeric: 1–5). We accounted for zero-inflation in the response. We assessed the interaction between environmental and nest size category but this was non-significant (p = 0.8) and removed from the presented model.

We evaluated if seeing many pollinators affected the feeling of having established a successful flower intervention (meadow or planting) using GLMs, with a binomial error distribution. We modeled the proportion of participants that stated they observed “many insects” (as opposed to “none, or few”), as a function of perceived intervention success.

We evaluated the accuracy of the simple pollinator assessments using a GLM with a negative binomial distribution, modeling meadows (N = 165) and plantings (N = 218) separately. We summed counts of the three major pollinator groups counted during surveys in 2021 (bees and wasps, butterflies, and hoverflies) to assess how pollinator abundance related to participants simple scores of insect abundance (“many insects,” as opposed to “none, or few”).

All analyses were carried out in R v 4.1.1 (R Core Team, 2021). Model assumptions were checked with package DHARMa (Hartig, 2020). Variance inflation factors (VIF) for models of meadows and plantings were checked with package car function vif (Fox and Weisberg, 2019), while VIF for bee hotels were checked with package performance (Lüdecke et al., 2021). Contrasts between groups (for example different sizes of plantings) were analyzed using Tukey's test for post-hoc analysis in the Emmeans package (Lenth, 2022). Test results were obtained from Analysis of Deviance Table using Wald Chi-square tests (package car function Anova, Fox and Weisberg, 2019) and the ggplot2 package was used to visualize data and create graphs (Wickham, 2016).

Respondents' gender, year of birth, and engagement in issues related to biodiversity (self-rated, Likert scale 1–5) were compiled to describe whom the campaigned had reached and involved. All respondents from the meadow- and flower-planting surveys were included, even if their answers had previously been removed from analyses of ecological questions due to incomplete data. We were interested in how changes to vegetation quality affects peoples' perception of their garden, and therefore did not include data from bee hotel-respondents. Moreover, in contrast to added flower resources, the benefits of bee hotels are contested.

The number of participants who saw many pollinators was related to the number of flowering species in the meadow (χ² = 40.247, df = 4, p < 0.001). Post-hoc tests revealed that fewer participants had reported many pollinators in meadows that contained 1–5, as compared to >5, flowering plant species (Figure 1). The meadow size and its surrounding environment had no significant effect on the likelihood of reporting many pollinators (size: χ² = 6.899, df = 4, p = 0.141; environment: χ² = 2.057, df = 3, p = 0.561), while meadow age showed a non-significant positive trend (age: χ² = 3.122, df = 1, p = 0.077). Meadow age and plant species richness were positively correlated (rho = 0.24, p < 0.001), but not strong enough to preclude inclusion in the same models (checked with VIFs, as above).

The proportion of participants that saw many pollinators was positively related to the age and size of the planting (age: χ² = 9.35, df = 1, p = 0.002; size: χ² = 31.24, df = 2, p < 0.001; Figures 2A, B). A lower proportion of participants saw many pollinators when plantings were situated in dense urban, compared to the other environments (χ² = 19.08, df = 3, p < 0.001, Figure 2C). The number of flowering plant species had no significant effect on the abundance of pollinators seen (χ² = 2.17, df = 4, p = 0.71). As for meadows, age and plant species richness were positively correlated (rho = 0.36, p < 0.001), but not strong enough to preclude inclusion in the same models (checked with VIFs, as above).

A higher proportion of participants who reported having seen many pollinators also viewed their meadow as successfully established (χ² = 108.46, df = 2, p < 0.001, Figure 3A). More participants reported they had observed many pollinators in meadows with the highest success rating, compared to meadows of either medium or low success, while participants with a medium success rate were in between the low and highly successful. Similarly, a higher proportion of participants who viewed their plantation as succesfull reported to have seen many insects, compared to those with a low or medium success rating (χ² = 87.85, df = 2, p < 0.001, Figure 3B).

Bee hotel occupancy rates were positively related to both local flower availability (χ² = 20.5, p < 0.001, Figure 4A) and bee hotel age (χ² = 69.85, p < 0.001, Figure 4B). Furthermore, occupancy rates differed between nest size categories (χ² = 101.99, p < 0.001, Figure 4C). Occupancy was significantly higher in small and medium sized holes (2–5, 6–10 mm diameter), than in large holes (11–15 mm). Occupancy rates differed significantly between environment types (χ² = 74.69, p < 0.001, Figure 4D). In particular, occupancy was significantly higher in both rural environments than either of the urban environments.

For both meadows and plantings, the pollinator abundance assess by 10-min surveys was highly positively related to if participants reported having seen “none or few” or “many” insects (meadows: z = 3.668, p < 0.001; plantings: z = 5.167, p < 0.001).

The vast majority (80%) of survey respondents were women and most (47%) were aged 41–60 years, while the categories aged 20–40 and 61–80 years made up 27% and 25%, respectively. The majority (69%) considered themselves to have a strong engagement (4–5 on a 1–5 Likert scale) in issues related to biodiversity conservation, while 26% classified themselves as equally committed as the societal average (3).

Positive effects of local flower species richness and abundance on the diversity and abundance of pollinator communities has previously been reported using traditional research methods, both in urban gardens (Quistberg et al., 2016; Del Toro and Ribbons, 2020), in rural experimental (Hegland and Boeke, 2006; Ebeling et al., 2008), and in agricultural settings (Potts et al., 2009; Jönsson et al., 2015). In addition, a citizen science project using standardized sampling methods showed that sown garden meadows enhanced local pollinator abundance and diversity over a 2-year period (Griffiths-Lee et al., 2022). Our results corroborate these findings, and in addition show that a very simple measurement, such as perceived pollinator abundance estimated by citizen scientists, may be used as a proxy for abundance to assess and compare the value of pollinator enhancement interventions.

The availability of local flower resources has been highlighted as a key factor for urban pollinator abundance and diversity (e.g., reviewed by Wenzel et al., 2019; but see Gathof et al., 2022) and may even buffer bee populations against the negative effects of landscape scale urbanization (Burdine and McCluney, 2019). Although we cannot evaluate the effects on pollinator populations in the wider landscape, even small-scale flower enhancements may result in population level effects if implemented on a large enough scale. For example, based on research in agricultural landscapes (e.g., Cong et al., 2014; Jönsson et al., 2015), one may expect that neighborhoods where uptake of interventions is high can support more pollinators at the landscape scale.

There was no effect of plant species richness in plantings on pollinator abundance. This may be because flowerbeds in general are highly dominated by ornamental and non-native plant species (Loram et al., 2007; Lowenstein and Minor, 2016), and therefore mainly cater for generalist pollinator species (Corbet et al., 2001; Wenzel et al., 2019). Adding more plant species to a flowerbed may then still only benefit the same part of the pollinator community and thus to a lesser degree attract more pollinators of other species (but see Simao et al., 2018; Staab et al., 2020). Simao et al. (2018) also show that, for small generalist bees, additions of urban flower resources had the strongest (positive) effect at low surrounding resource levels, whereas at higher levels the effect was unpredictable. Most (75%) of respondents reported plantings from single-family housing areas and we expect a generally high level of flowering of ornamental plants in such locations. In contrast, adding more plant species to a garden meadow dominated by native plants may increase the attractiveness of the garden to a wider array of pollinator species, including some specialists. Participants were only asked about the number of plant species present in their meadows or plantings, not about the abundance. It is therefore possible that we had seen a positive effect of flower abundance on pollinator activity in both meadows and plantings (as we did for occupancy of bee hotels), had we measured this variable. Indeed, the size of plantings, which is likely positively related to flower abundance, had a positive effect on pollinator abundance.

Our results highlight the enhanced benefit of older flowering elements and bee hotels, and thus the need for gardeners to make more lasting commitments to changes in garden design and management. The significant positive effect of planting age (and the non-significant positive trend for meadows) could be because perennial plants, which are often preferred by bumblebees (Fussell and Corbet, 1992), often require several years to establish and flower from seed. Gardeners may make several attempts at sowing or planting new species into an intervention, thus intentionally increasing plant richness over time. In addition, spontaneous establishment of plant species, especially in garden meadows, may lead to increased plant diversity over time (Norton et al., 2019). Age and plant diversity were indeed positively correlated. Another explanation could be that beneficial micro-habitats build up over time in gardens and flower beds, including bare patches of soil for ground nesting bees, dead wood and stems with hollows for cavity nesters, and dead organic matter for some hoverfly taxa, allowing a delayed response of pollinator populations to an intervention. For bee hotels, the philopatric behavior of many solitary bee species may explain why occupancy builds up over time. The increased occupancy of older nests has previously been described for the common species Osmia bicornis (Steffan-Dewenter and Schiele, 2004; Fortel et al., 2016).

The largest nest cavities (11–15 mm) were far less inhabited than smaller ones (2–5 and 6–10 mm). Most likely, this is due to there being few bee or wasp species in Sweden that use nests larger than 10 mm; Recommendations for bee hotels in temperate regions rarely stretch past 12 mm (e.g., Naturhistoriska Riksmuseet, 2013; Bauer et al., 2015; Winter, 2018). Clear information about preferred size and design of bee hotels may thus increase the occupancy of hotels in future campaigns.

Bee species differ in their requirements for nesting conditions. Of Sweden's approximately 250 solitary bee species, around 70% are ground nesters, and only a small fraction of species are known to nest in bee hotels (Linowski et al., 2004; Naturhistoriska Riksmuseet, 2013). Despite this, bee hotels may be useful bio-indicators for insect pollinators in general (Tscharntke et al., 1998). A garden where a bee hotel is highly occupied can thus be expected to host many other pollinating insects, either nesting in and/or visiting the garden to forage.

We found that the surrounding environment moderated pollinator abundances in flower interventions, such that plantings in single-family urban or rural gardens and rural natural environments were more visited by pollinators, compared to yards and green spaces in dense urban areas. Similarly, bee hotels in both rural gardens and natural environments were more occupied compared to those in both types of urban sites. Our results thus corroborate previous research showing that urbanization is generally negative for insect abundance and diversity, including pollinators (Fortel et al., 2014; Geslin et al., 2016; Fenoglio et al., 2020; Piano et al., 2020). However, pollinator taxa and trait groups differ in sensitivity to urbanization. Butterflies (Fenoglio et al., 2020; Piano et al., 2020) and hoverflies (e.g., Verboven et al., 2014; Persson et al., 2020) are generally negatively affected by urbanization and, while a recent meta-analysis show that bee diversity is negatively affected by urbanization (Fenoglio et al., 2020), other studies have shown that cavity nesting and long tongued bee species may actually benefit from intermediate to high levels of urbanization (Fortel et al., 2014; Wenzel et al., 2019). Our results show that urban bee hotels were less occupied than those in rural sites, indicating that cavity nesting species were actually less abundant in urban areas of Sweden. This could partly be explained by the large geographical uptake of the campaign, whereby we likely included rural sites spanning from those embedded in production landscapes to those rich in semi-natural or natural habitats, where the latter may harbor high bee abundance and diversity. Alternatively, and a bit speculative, bee hotels in urban areas may be of lower quality than those in rural or natural sites; e.g., they may more often be store-bought rather than home-made or place-built, and/or placed in too exposed or too shaded sites. This may make them less attractive to nesting bees compared to those in rural/natural sites (von Königslöw et al., 2019).

Regarding bee body size, results are so far inconclusive. While some studies show that small bodied species may benefit from highly urbanized areas (Banaszak-Cibicka and Zmihorski, 2012; Gathof et al., 2022), other studies find the opposite (reviewed by Wenzel et al., 2019). In addition, body size and nesting substrate may be correlated, such that small bees more often are ground nesters (Banaszak-Cibicka and Zmihorski, 2012). Our results do not indicate that any size class of cavity nesting bees benefit from urban areas (non-significant interaction for environment and size class). However, we have not assessed ground nesting species and may therefore have missed genera that are particularly well-adapted to urban environments.

Flowering interventions were seen as more successful when they attracted many pollinators, indicating that respondents evaluated their interventions based on the desired ecological outcome (to provide flowers and benefit pollinators). Previous research has shown that pro-pollinator actions may be conditional on the degree to which people perceive that their actions will indeed benefit pollinators (Knapp et al., 2021). Although not tested here, this may lead to a reinforcing loop, where perceived successful interventions remain, while less successful ones are terminated.

People who choose to design and manage their gardens to benefit biodiversity do so for a multitude of reasons, ranging from aesthetics and personal well-being to a sense of moral responsibility for nature (Freeman et al., 2012; Goddard et al., 2013; Knapp et al., 2021). People who are personally engaged and interested in biodiversity may also be more likely to perform acts beneficial to biodiversity (see e.g., Maiteny, 2002). This could explain why the majority of the respondents in this study stated that they are highly engaged in issues concerning biodiversity. Indeed, a growing number of studies highlight citizen science as a tool in the transition to a more sustainable society by strengthening, encouraging, and validating public participation in environmental and sustainability issues (Dickinson et al., 2012; Shulla et al., 2020).