Determinants of honeybee hive survival within cities and its use to monitor urban biodiversity: a Canadian study

doi:10.21203/rs.3.rs-2973303/v1

Cities are shown to provide favourable conditions for western honeybees (Apis mellifera) by protecting them from agricultural pesticides and offering a greater diversity of flora. Nevertheless, bee populations continue to decline in many urban areas, posing threats to plant reproduction and urban biodiversity, especially as many now promote urban beekeeping as part of their conservation efforts. While current research primarily focuses on the causes of bee colony collapse, including pesticide exposure, climate change, and habitat fragmentation, little attention is dedicated to urban apiculture and factors associated with the survival of honeybees within cities. Here, we analyse data from 3,694 honeybee hives in Montreal and Toronto to assess the effect of urban and environmental factors on hive survival. We find ground-level ozone concentrations, elevation, and the number of surrounding hives to be negatively associated with hive survival, which may point towards an issue of hive saturation. Conversely, vegetation density and the COVID-19 pandemic are positively associated with the likelihood of hive survival, emphasizing the effect of urban air pollution. These findings highlight the need for urban planners to consider neighborhood-scale environmental factors to support conservation efforts and identify potential venues for enhancing bee survival within cities beyond existing urban beekeeping initiatives.

Biological sciences/Ecology/Biodiversity

Biological sciences/Ecology/Urban ecology

As global populations of bees and other pollinators continue to dwindle, many cities are seeing an uptake in urban beekeeping. This movement is occurring as many cities embrace conservation planning strategies and increasingly attempt to draw on the inherent benefits of nature. The Western honeybee (Apis mellifera) and other pollinators appear to be well suited for the urban landscape as it is believed to shield them from agricultural pesticides and predators, offer a greater diversity of flora, and even extend the pollinating season by providing warmer surface temperatures and wind protection^1,2. These ostensible benefits are corroborated by recent research that finds a higher bee diversity and lower rates of beehive mortality within cities in comparison to their rural counterparts^3,4. That said, not all researchers agree with some pointing to poor urban air quality to dismiss urban environments as favourable for bees^5,6, and to date a consensus has largely not yet been achieved. Nevertheless, despite the potential advantages that the urban environment may present, cities are not immune to bee colony collapse and bee populations have continued to decline at an alarming rate within the urban context. This is the case in Canadian cities, where urban beekeepers experienced above average hive mortality rates and the losses were further compounded by importation restrictions due to the COVID-19 pandemic⁷.

The bee population decline is worrisome as 87.5% of flowering plants depend on animal vectors for pollination³, and a higher mortality rate of bees within cities could have severe repercussions on urban biodiversity. As more cities recognize the importance of preserving their biodiversity and start to allocate resources toward this commendable goal, an understanding of the factors that contribute to the health and survival of bees within urban areas will undoubtedly prove helpful, especially given the synergies and role such pollinators play in safeguarding our biodiversity and sustaining our urban ecosystems.

While extensive research has explored the causes of the ongoing bee colony collapse, such as pesticide exposure^8,9, climate change^10,11, and habitat loss^12,13, relatively little attention has been directed to urban apiculture and to understanding how the urban form and environment may influence the health and survival of honeybees.

Results from studies having previously compared the abundance, health, and diversity of bees within urban and rural areas remain largely inconclusive, as some find urban areas to contain lower levels of pesticides, offer more diverse flora, and provide floral resources throughout the seasons, which together allow for a temporally longer and richer habitat for bees^1,3,6,14,15. However, others disagree and point towards poor urban air quality, the presence of surface and water contaminates, and the saturation of existing bee colonies within cities as reasons to dismiss urban environments as favourable for bees^5,6. Despite remaining inconclusive, what is clear from this existing body of work is that urban areas must not be overlooked as a habitat for pollinators and that the abundance, health, and diversity of bees, which have been shown to vary significantly between urban and rural areas, may also vary within cities themselves.

The unique combination of land use, urban form, and environmental attributes that characterise each part of a city influences levels of pollution and vegetation and may ultimately determine the health and survival of honeybees. Wealthier neighbourhoods, for instance, are often shown to contain more green space^16,17, whereas poorer parts of the city are commonly exposed to higher levels of pollution due to their proximity to highways and industries^18,19. The presence and size of gardens are also found to be positively correlated with higher levels of homeownership and the proportion of detached and semi-detached housing in a given area^20,21, which in turn may influence the availability and diversity of surrounding flora. Such urban characteristics have already been shown to affect pollinators, as illustrated in research conducted in Chicago which found bee abundance to be correlated with population density²² and similar results have also been established in St. Louis, where bee diversity was found to be associated with population density and household income²³. Given the mosaic of buildings, streets and green spaces that characterize urban landscapes, and the growing body of evidence to suggest that such attributes may influence the health bees, it is imperative to analyze this relationship at a much finer scale and assess which characteristics of the urban form and environment affect the survival of honeybee colonies within cities.

Here we use data from 3,694 hives in Montreal and Toronto to conduct a series of logistic regressions and determine the factors that influence the survival of honeybee hives within urban areas. We find that lower levels of pollution and higher homeownership rates in the area surrounding the hive increase its chances of survival. Having more neighbouring hives in a 1-kilometre radius from the hive is further found to negatively affect its chances of survival, and possibly points to a problem of hive saturation within areas of these cities. Based on our results, we further speculate that the allowance of pesticides within city limits may have negatively affected the likelihood of hives survival, as hives in Toronto – where pesticides were banned for the entire duration of our study – fared significantly better than those in Montreal. Lastly, having data spanning from 2017–2022 enabled us to assess the impact of the COVID-19 pandemic, which appears to have had a positive effect on the health of honeybee hives in both cities. We surmise this may be due to lower levels of air pollution as vehicle traffic ground to a halt. These results highlight the importance of considering urban and environmental attributes when designing cities as a refuge for insect pollinators and demonstrate the potential for enhancing urban biodiversity through the promotion of bee health in cities.

To assess the effect of urban and environmental factors on the likelihood of honeybee hive survival within cities, we use data on the status (e.g., alive or dead) of hives belonging to Alvéole, an urban beekeeping company based out of Montreal. Out of the 3,694 hives used in this analysis from 2017-2022, we observe a survival rate of 69.2%. Figure 1 displays the density of Alvéole’s hives in both Montreal and Toronto during our study period.

Our analysis considers socioeconomic attributes as well as elements of the built form and environment within a 1-kilometre radius of the beehives that may influence their survival in urban areas. This radius represents the typical foraging range of western honeybees, encapsulating the availability and quality of foraging resources in the surrounding landscape^24,25,26. Figure 2 provides an example of how such variables were collected using the Normalized Difference Vegetation Index (NDVI) values within the 1-kilometers buffers from the hives. Table 1 presents the explanatory variables descriptively and a more detailed description, including the source and measurement of these variables can be found in the Methods section.

Table 1. Descriptive Table.

Continuous variables	Variable description	Min	Max	Mean	Std. Dev.
NDVI_Low	Percentage of a buffer's surface area containing NDVI values of 0 to 0.2	0.33	0.97	0.60	0.24
NDVI_Medium	Percentage of a buffer's surface area containing NDVI values of 0.2 to 0.5	0.03	0.92	0.36	0.21
NDVI_High	Percentage of a buffer's surface area containing NDVI values of 0.5 to 1	0.00	0.51	0.03	0.06
Water	Percentage of a buffer's surface area containing watercourses or waterbodies	0.00	0.69	0.04	0.10
Highway length	Length of highways (metres) within buffer	0	12 211	2 858	2 749
Smoke	Annual average concentration of PM2.5 (ug/m3)	6.57	7.13	6.91	0.15
O3	Annual average ground-level ozone concentrations (ppb)	22.09	27.34	23.54	1.46
Ownership rate	Mean property ownership rate within buffer	0.18	0.98	0.48	0.17
Hive Count	Number of neighboring hives within buffer, including observed hive	1.00	192.00	31.06	40.26
Categorical variables	Category measured	Count		Percentage
Hive Status	Alive	2551		69.1 %
Hive Status	Dead	1143		30.9 %
Water	Containing less than the mean area of water within all buffers in our sample	2869		77.7 %
Water	Containing at least as much as the mean area of water within all buffers in our sample	825		22.3 %
Highway length	Containing less than the mean area of highways within all buffers of our sample	1827		49.5 %
Highway length	Containing at least as much as the mean area of highways within all buffers of our sample	1867		50.5 %
Placement	Ground (hives located below 3 floors)	1060		28.7 %
Placement	Roof (hives located 3 floors or above)	2634		71.3 %
City	Montreal	2208		59.8 %
City	Toronto	1486		40.2 %
Hive Count	0-2	502		13.6%
	3-5	684		18.5%
	6-10	660		17.9%
	11-25	650		17.6%
	26-75	556		15.1%
	75-192	642		17.4%

Overall, results established using the reduced model, obtained through a backward stepwise selection process (See Table 2), indicate that the surrounding level of vegetation, air quality, number of neighboring hives and height of the hives themselves significantly influence their likelihood of survival within urban environments. We also find that hives in Toronto have a significantly higher probability of survival. We accounted for the possibility of city-specific differences as our analysis of the spatial clustering of vegetation density (measured through NDVI values) revealed that compared to Montreal, Toronto’s vegetation is more dispersed. While high NDVI values appear to be spatially clustered in both cities, they are more clustered in Montreal (Moran’s I > 0.73, p <0.001) than in Toronto (Moran’s I > 0.56, p < 0.001). To account for this, we included an interactive Toronto*NDVI_High variable along with the NDVI and City dummy main effect variables.

The odds-ratios of NDVI variables show that levels of vegetation have a strong effect on the probability of hive survival. A 1% increase of Medium NDVI, previously found to represent shrub and grass within cities²⁷, is associated with 45.4% lower odds of survival. Conversely, a 1% increase in High NDVI, which have been categorised as urban forests²⁷, is associated with 8.4 higher chance of survival. While our analysis is conducted using 1-kilometre buffers, as this is shown to be the typical foraging distance of honeybees within cities^24,25, we also replicate the models using 3-kilometre buffers. This tests the robustness of the results and accounts for situations where surrounding levels of vegetation are sparse, and bees may need to travel further distances to satisfy their foraging needs. All variables. Including NDVI, retain their sign and significance using the larger buffers.

Our models reveal that higher annual average ground-level ozone (O3) concentrations negatively affect hive health. A 1 part per billion (ppb) increase in ground-level O3 concentration at the postal code nearest to the hive corresponds to a 10% decrease in its chances of survival. Note that the daily O3 ground-level concentration variance ranges, in the case of Toronto, from 5-40 ppb²⁸, thus accentuating its overall effect.

Results of the above continuous variables were analysed using odds ratios as they exhibit dispersed distributions and considering the effect at the mean may be misleading. To facilitate interpretation, results for the binary and categorical variables will now be presented in terms of marginal effects while holding all other variables at their means.

On average, hives located in Toronto have a 7% higher predicted probability of survival than those in Montreal. Since the interaction variable is not statistically significant, we may infer that Toronto’s higher survival rates are not associated to its more dispersed and/or abundant level of vegetation in proximity to the hives. Instead, this may be due to other city-specific attributes such as its average warmer temperatures enabling longer foraging season, and its bylaw restricting the application of pesticides on all public and private properties. Indeed, pesticides were banned in Toronto as of April 2004, whereas their usage has only been prohibited in Montreal as of January 1^st, 2022.

The number of neighbouring hives also plays a role on survival. Compared to hives without neighbours in their 1-kilometre buffer, those with neighbouring hives have a lower likelihood of survival, and this is found to steadily decline as the number of neighbouring hives increases; ranging from a 30% lower likelihood of survival when there are less than 10 neighbouring hives, to a 46% lower survival rate when there are more than 75 neighbouring hives within the 1-kilometre buffer). A similar, but less pronounced pattern emerges when considering the number of neighbouring hives within the 3-kilometre buffer. The more sizeable reduction in hive survival observed in the 1-kilometre analysis compared to the 3-kilometre analysis, may be explained by the higher proximity of hives and fiercer competition. This is especially true given that our hive saturation analysis only considers hives belonging to Alvéole and does not account for other beehives in the buffers (not managed by Alvéole), nor the presence of other pollinators such a bumblebees.

The elevation of the hive also influences its likelihood of survival, as hives located above three floors have on average an 8% lower probability of survival. Since our dataset spans from 2017-2022, this allows us to assess the impact of the COVID-19 pandemic. We find that compared to 2019, hives in 2020 and 2021 had a 11% and 15% higher likelihood of survival.

Table 2 Logistic regression results for predicting hive survival within cities.

	Model 1.1 (Full Model)				Model 1.2 (Reduced Model)
	Coeff. and sig.	Marginal effects	Odd Ratio	Confidence interval (95%)	Coeff. and sig.	Marginal effects	Odd Ratio	Confidence interval (95%)
City Montreal (ref.)
City Toronto	0.61 ***	0.12	1.84	(1.34, 2.53)	0.37 ***	0.07	1.44	(1.21, 1.73)
Toronto * NDVI_High	-2.67	-0.55	0.07	(0.0, 1.4)
NDVI_Medium	-0.75 *	-0.16	0.47	(0.25, 0.9)	-0.60 *	-0.12	0.55	(0.33, 0.92)
NDVI_High	2.65 **	0.55	14.14	(2.37, 84.29)	2.13 **	0.44	8.40	(2.0, 35.26)
Water (Binary)	-0.03	-0.01	0.97	(0.8, 1.19)
Highway length (Binary)	0.00	0.00	1.00	(0.85, 1.19)
Smoke	-0.78	-0.16	0.46	(0.18, 1.16)
O3	-0.13 ***	-0.03	0.88	(0.82, 0.94)	-0.11 ***	-0.02	0.90	(0.85, 0.95)
Ownership rate (Mean)	0.20	0.04	1.22	(0.5, 2.98)
Median Household Income 1st Quartile (ref.)
Median Household Income 2nd Quartile	0.06	0.01	1.06	(0.86, 1.31)
Median Household Income 3rd Quartile	0.05	0.01	1.05	(0.82, 1.34)
Median Household Income 4th Quartile	-0.02	0.00	0.98	(0.72, 1.33)
Placement ground (ref.)
Placement roof	-0.35 ***	-0.07	0.70	(0.58, 0.85)	-0.40 ***	-0.08	0.67	(0.56, 0.8)
Hive Count <2 (ref.)
Hive Count <5	-1.17 ***	-0.27	0.31	(0.23, 0.43)	-1.35 ***	-0.31	0.26	(0.18, 0.36)
Hive Count <10	-1.09 ***	-0.25	0.34	(0.24, 0.47)	-1.32 ***	-0.30	0.27	(0.19, 0.38)
Hive Count <25	-1.39 ***	-0.32	0.25	(0.18, 0.35)	-1.60 ***	-0.37	0.20	(0.14, 0.29)
Hive Count <75	-1.32 ***	-0.31	0.27	(0.18, 0.39)	-1.49 ***	-0.34	0.23	(0.15, 0.33)
Hive Count 75+	-1.87 ***	-0.43	0.15	(0.1, 0.23)	-2.02 ***	-0.46	0.13	(0.09, 0.2)
Year 2017	0.04	0.01	1.04	(0.74, 1.47)	0.09	0.02	1.09	(0.77, 1.54)
Year 2018	-0.21	-0.05	0.81		-0.23 *	-0.05	0.80	(0.64, 0.99)
Year 2019 (ref.)
Year 2020	0.59 ***	0.11	1.80	(1.41, 2.3)	0.61 ***	0.11	1.85	(1.44, 2.36)
Year 2021	0.80 ***	0.15	2.23	(1.74, 2.86)	0.81 ***	0.15	2.25	(1.75, 2.89)
Year 2022	0.21	0.04	1.23	(0.98, 1.54)	0.22	0.04	1.25	(0.99, 1.56)
Observations	3 694				3 665
Chi2	282.94				297.22
Significance	0.00				0.00
Pseudo R2 (McFadden)	0.06				0.07
Pseudo R2 (Naglekerke)	0.10				0.11
AIC	4335.5				4264.5
Notes: * p < 0.05; p < 0.01; * p < 0.001 Model 1.2 (Reduced Model) is obtained through a backward stepwise selection process.

Designing urban areas that support the survival of bees and promote biodiversity requires a thoughtful consideration of environmental factors surrounding beehives. First, our findings highlight the significance of vegetation density in the survival of beehives. Hives surrounded by high NDVI values (0.5-1) have a greater chance of survival whereas those with a larger proportion of medium NDVI values (0.2–0.5) in a 1-kilometre buffer are correlated with lower rates of survival. Previous studies have shown that urban areas offer more abundant and diverse flora such as flower gardens and parks compared to rural areas where intensive agricultural activities are more common^29,30. Given that well-planned urban greenery are also shown to support biodiversity and create sustainable habitats for bees^1,2, a straightforward policy recommendation would be to implement minimum green space requirements for new constructions.

Second, the negative association between ground-level ozone concentrations and hive survival rates in both Montreal and Toronto confirms previous research^31,32, emphasising the harmful effects of air pollution, particularly O3 exposure, stemming from vehicle emissions, power plants, industrial boilers, refineries, and other sources, on honeybee populations and their foraging behaviour^33,34. Previous studies have shown that acute O3 exposure may cause olfactory generalization in pollinators and impair honeybees’ ability to detect, learn and memorize floral odors²². We find that even smaller levels of exposure within an urban context may be problematic, and that each ppb increase in O3 levels in the area surrounding the hive can be associated with a significantly lower likelihood of it surviving. The negative correlation between O3 levels and hive survival rates further highlights the need to address air pollution through policies promoting cleaner transportation, eco-friendly industry practices, and environmentally conscious activities. To address this issue, urban planners could also promote the development of green space corridors connecting various natural habitats within the city. These corridors would facilitate honeybee foraging and reduce their exposure to air pollution. An example of this is the Darlington ecological corridor currently underway in Montreal, which aims to create a green link between university campuses and other neighbourhood areas in the city³⁵.

Third, our results indicate that the survival rate of hives in our sample improved at the onset of the COVID-19 pandemic. As travel ground to a halt, so did levels of vehicle emissions and cities around the globe saw a considerable decline in their rate of urban air pollution³⁶. Not only did the pandemic improve urban air quality, but it also enhanced vegetation growth and may ultimately have improved the well-being of honeybees^37,38. Both Toronto and Montreal saw a stark decline in traffic levels during the COVID-19 pandemic³⁹, which corroborates the effect of air pollution and further accentuates the need to promote cleaner forms of transportation to enhance urban biodiversity.

Fourth, our research found that the elevation of hives is associated to their survival, as hives located on 3-story roofs or higher have a significantly lower likelihood of survival. This finding challenges the belief that beekeeping is equally successful at any urban location and emphasises the need for strategic hive placement to maximise their chances of survival. Previous studies have also stressed the importance of hive height, particularly in urban settings, where increased altitude leads to higher wind levels, making foraging more challenging for bees^40,41. MacIvor⁴¹ found that in Toronto, hives situated on rooftops of five stories or higher had lower survival rates compared to lower-altitude hives, possibly due to more extreme weather conditions at greater heights. Beekeeping companies, such as Alvéole, are already addressing this issue by usually not installing hives above ten stories. Recognizing that adverse effects can occur even below this threshold could help further improve bee health and hive survival. To minimize the negative impacts of elevation on existing honeybee colonies, cities should encourage the development of green walls, green roofs, and other measures to add vegetation at varying altitudes within cities. Moreover, a policy recommendation going forwards would be to limit the height at which honeybee hives may be located to avoid having them beyond a certain designated height.

Fifth, our study further underscores the need to take the environmental carrying capacity into account and implement efficient hive density management. The presence and sheer quantity of neighbouring hives is shown to be negatively correlated to the likelihood of hive survival. This finding concurs with research by Iwasaki & Hogendoorn⁴², highlighting the importance of hive density management in urban settings to prevent resource competition and disease transmission. Urban beekeeping is on the rise and the current lack of regulations surrounding hive saturation or even information concerning a sustainable level of hive density may have negative consequences on urban biodiversity and harm pollinators including honeybees themselves. To address this issue, a thoughtful approach would require collaborating with beekeepers, environmentalists, and members of the community to identify specific locations for urban beekeeping. To further reduce resource competition and promote the well-being of honeybees within cities, policymakers should consider limiting the number of hives within their jurisdiction. This would entail further research to determine the carrying capacity of cities and establish an acceptable hive density threshold level.

Sixth, we have observed a higher hive survival rate in Toronto compared to Montreal. As mentioned earlier, this may be due to the former having higher levels and density of vegetation and tree coverage than Montreal, but it may also be due to Toronto prohibiting the use of pesticides which are proven to increase honeybee colony mortality^43,44, Montreal has also since banned the use of pesticides “to protect the health of the population, preserve the environment and biodiversity”³⁵. However, as the ban only took effect on January 1st, 2022, and our data collection largely concluded in June 2022, its benefits may not yet have fully materialised. This in turn, may explain why we do not find hives in 2022 to have a significantly higher survival rate than those in 2019. This could also be explained by the fact that in 2022 many pandemic-related restrictions were lifted and vehicle air pollution had largely returned to pre-pandemic levels. Nevertheless, more research is needed to measure whether the gap in hive survival between Toronto and Montreal will decrease in response to the latter also prohibiting the use of pesticides.

To conclude, our research adds a new perspective to the existing literature by complementing the focus on macro (e.g., climate change) and micro (e.g., parasites) ecological factors with a neighbourhood scale analysis, highlighting the importance of the urban form and environment when considering the survival of honeybees within cities. For instance, we have shown that the likelihood of hive survival is associated with its elevation and level of surrounding vegetation, both of which are a function of the built environment. Our results call for targeted strategies to improve the local environment around beehives and suggests new avenues for further research to improve the survival of beehives as part of broader biodiversity conservation efforts.

Source and measurement of variables

To examine the factors that influence the survival of beehives within urban areas, we use data from Alvéole, an urban beekeeping company based out of Montreal, which possesses hives in over 40 cities across North America. It includes information on all honeybee hives belonging to the company in Montreal and Toronto between 2017 and 2022. The data is compiled and updated by Alvéole employees upon each visit to the hives and contains the date of each hive visit, the status of the hive (e.g., alive or dead), the cause of death (if the hive is found to be dead), the geographical coordinates of the hive, and a variable to help us estimate its elevation. The status of the hive is used as the binary response variable in our models (1 = Alive; 0 = Dead). Beyond removing duplicates and null observations during data cleaning processes, we also removed entries with irrelevant causes of death including animal/human interference such as vandalism or robbing, poor beekeeping practices, and swarms as these may have biased our results. A total of 3,694 hives were retained in our dataset over the entire study period.

Using QGIS’s geoprocessing tools to create buffers around the hives, we are able to consider elements of the built form and environment within a 1-kilometre radius of the beehives that may influence their survival in urban areas. This radius represents the typical foraging range of western honeybees, encapsulating the availability and quality of foraging resources in the surrounding landscape^24,25,26. To ensure the robustness of our results a similar approach is employed to create 3-kilometre buffers around the hives, which may better capture the maximum distance honeybees are willing to travel to forage within cities^45,46.

The source and measurement of explanatory variables vary quite significantly, and we will divide them into three categories: Environmental attributes, neighbourhood attributes, and additional hive attributes.

Environmental attributes

NDVI values were used to assess the density of vegetation in the area surrounding the hives. Landsat 8 data were obtained at a 30m resolution from June 2018. The specific year and date were selected to minimize imagery with cloud coverage and to provide a more accurate depiction of vegetation density levels. NDVI values were overlaid over our study locations (City limits of Toronto and Montreal) and subsequently divided into three categories (NDVI values between 0-0.2; NDVI values between 0.2–0.5; NDVI values between 0.5-1.0), based on previously established urban vegetation classes and their related NDVI values²⁷. These categories were abbreviated to NDVI_Low; NDVI_Medium; and NDVI_High in our models. The NDVI_Low category also includes values inferior to 0 and corresponds to surface barren of vegetation such as roads, buildings, and bodies of water. To determine the density of vegetation surrounding each hive, the percentage of pixels corresponding to each NDVI category within the 1-kilometre buffer was retained.

Access to water was also included in our models and was measured by calculating the area of water within the 1-kilometre buffer. Specifically, we queried OpenStreetMaps for bodies of water and used available hydrography and watercourses polygon features available on the City of Montreal and Toronto websites to calculate the area of water within the 1-kilometre buffers surrounding the hives. To simplify, we then created a binary variable using the mean area of water within the buffers to include in our models.

Air pollution was assessed through annual average ground-level ozone (O3) concentration (parts per billion) and annual average concentration (ug/m3) of fine particulate matter (PM2.5). The PM2.5 and O3 variables were provided by the Canadian Urban Environmental Health Research Consortium (CANUE). The average concentration of PM2.5 was estimated by combining the aerosol optical depth retrievals from the NASA MODIS, MISR, and SeaWIFS instruments⁴⁷. Ground-level ozone concentrations were calculated using the Global Environmental Multi-scale Modelling Air Quality and Chemistry (GEM-MACH) model monitored by Environment and Climate Change Canada⁴⁸. More information on how this data was calculated can be found on the CANUE and ECCC websites. The obtained values were spatially segmented into square-shaped pixels and joined to the hives’ nearest postal code using geoprocessing tools and the polygons’ centroid.

Neighbourhood attributes

Homeownership and median household income were included in the analyses as they have previously been linked to the presence of gardens in urban areas^20,49 and may, in turn, indicate a wider diversity of flora⁵⁰. Homeownership and median household income data were collected from the 2016 Census of the Population⁵¹. The percentage of homeownership was calculated at the dissemination area and the mean percentage of homeownership of all dissemination areas comprised within the 1-kilometre buffer around the hives was retained. A natural neighbours interpolation approach was employed⁵², to account for the smooth nature of homeownership and the non-uniform distribution of our data. This variable is referred too as Own_mean in our models. A similar process was employed for median household income, but we chose to subsequently transform this variable into categories using quartiles to determine whether a certain income threshold influenced honeybee survival. City-specific cutoffs were used to create quartile categories, meaning that the 1st quartile range for Toronto is higher than that of Montreal, based on the city's overall household income distribution.

Previous studies have shown that the surrounding density and proximity of roads are associated with higher levels of metals (e.g., lead, zinc, cadmium) in insects, which in turn may reduce bee diversity and abundance⁵³. Anthropogenic noise, such as that produced by vehicle traffic, may also trigger behavioural changes in animals⁵⁴, and to account for this we included a variable in our models to capture the length of highways within the 1-kilometre radius from the hives. This data was obtained from the websites of the Communauté métropolitaine de Montréal⁵⁵ and the Ontario Ministry of Natural Resources and Forestry⁵⁶.

Additional hive attributes

To account for city-specific differences between Montreal and Toronto, namely the fact that Toronto prohibits the use of pesticides, whereas Montreal only banned them on January 1st, 2022. We include a dummy variable to account for the city in which the hive is located. Furthermore, to capture the potential effect of the COVID-19 pandemic, such as the decrease in vehicle traffic³⁹, we also include a year dummy variable.

Alvéole’s dataset also provides information on the elevation of hives. Specifically, the PlaceRoof variable tells us whether the hive is located on the ground or on a low roof (below 3 floors), or whether it is situated at a higher elevation (3–10 floors). Knowing that elevation may play a role, Alvéole does not install hives on roofs higher than 10 floors, which is why the variable is capped at that level.

Having access to all of Alvéole’s hives in both Montreal and Toronto, we were also able to account for the density of surrounding hives within the 1-kilometre buffer. In order to establish an acceptable threshold and help implement efficient hive density management, we transformed the number of surrounding hives into 6 categories using sextiles and adjusting slightly to the closest round number to facilitate interpretation and eventual policy use. A breakdown of the surrounding hive count categories, as well as details on the distribution of other explanatory variables, can be found in Table 1.

Model specifications

A series of logistic regression models were used to understand the factors that influence the survival of beehives within urban areas. This form of regression model is well suited to handle dichotomous dependent variables and is effective in predicting the presence or absence of an outcome based on a set of predictor variables. Our response variable is the hive status, which is dichotomous by nature and either takes a value of 1 (if alive), or 0 (if dead). The likelihood of a hive surviving can be expressed in terms of probabilities, as shown in Eq. 1.

$$P\left({Y}_{1}=1\right) = {\pi }_{i}$$

(1)

$$P\left({Y}_{1}=0\right) = {1-\pi }_{i}$$

Equation 2 illustrates how these probabilities can in turn be expressed in relation to a series of predictors (${X}_{1},{X}_{2}, \dots , {X}_{n})$ and the logistic model form can be reached using a logit transformation (see Eq. 2).

$$P(Y=1|{X}_{1},{X}_{2}, \dots , {X}_{n})$$

(2)

$$ln\left(\frac{{\pi }_{i}}{1-{\pi }_{i}}\right)= \alpha + {\beta }_{1}{X}_{1}+ {\beta }_{2}{X}_{2}+\dots + {\beta }_{n}{X}_{n}$$

Where ${\pi }_{i}$is the probability of the outcome of interest, $\alpha$ is the intercept, and $\beta$ are the regression coefficients. We present the coefficients of the full and reduced models, obtained through a backwards stepwise selection process. All model coefficients, marginal effects, odd ratios and confidence intervals are displayed in Table 2. Model fit is assessed using both the McFadden’s pseudo-R2 and Nagelkerke’s pseudo-R2. All statistical procedures were performed in R (Version 4.2.1).

Model Assumptions

A critical assumption of logistic regressions is linearity between the log-odds of the outcome and each continuous independent variable. To ensure we conform with this assumption, we conduct the Box-Tidwell test, available in R under the car library⁵⁷. This test adds log-transformed interaction terms between continuous independent variables and their corresponding natural log into the regression model. Continuous independent variables that are found to be significantly related to their interaction terms were subsequently transformed into categorical variables to satisfy this assumption.

The time-series nature of our dataset also raises the risk of observation being influenced or related to other observations. Fortunately, this issue was avoided given the lifespan of Western honeybees and the fact that most of the hives in our dataset were not present throughout the entire study period. That said, the deviance residuals were plotted against the index numbers of the observations to visually ensure no serial correlation was present.

Logistic regression further assumes that highly influential outliers have been removed from the model as these may distort the accuracy and outcome of the model. To identify influential data points and subsequently remove them, we use Cook’s Distance using an 8/N cut-off threshold.

To ensure variables are not highly correlated we use the variance inflation factor (VIF) and in accordance with Akinwande et al.⁵⁸ consider values above 5 to indicate a problematic amount of multicollinearity. We further test for spatial autocorrelation to ensure the model’s residuals are independent from each other using the Moran’s I test, available in the spdep library in R⁵⁹. Results of the Moran’s I test (Moran’s I > 0.006, p < 0.133) indicate that there is no spatial autocorrelation and we do not need to account for the spatial arrangement of our data.

Availability of data and materials

All data needed to evaluate the conclusions in the paper are made available on the author’s GitHub repository at https://github.com/Gavin-MacG/Hive.git. R code used to generate results are also available upon request.

Acknowledgements

The authors would like to thank Alvéole, and particularly Alexandre McLean, for giving us access to their hive data and providing valuable feedback on the manuscript. Additional data used for this research were provided by the Canadian Urban Environmental Health Research Consortium (CANUE). We further wish to acknowledge financial support received from the Social Sciences and Humanities Research Council of Canada (SSHRC). The contents of this article reflect the views of the authors, who are responsible for the accuracy of the information presented.

Authors’ contributions

M.Y. and G.M. were involved in the collection of the built form and environmental data. M.Y., G.M. and G.A.T. analyzed the data. G.M. prepared Figures 1-2. All authors wrote and reviewed the manuscript.

Competing interests

The authors declare that they have no competing interests.

Baldock, K. C. et al. Where is the UK's pollinator biodiversity? The importance of urban areas for flower-visiting insects. Proceedings of the Royal Society B: Biological Sciences, 282, 20142849; https://doi.org/10.1098/rspb.2014.2849 (2015).
Hall, D. M. et al. The city as a refuge for insect pollinators. Conservation Biology, 31, 24-29; https://doi.org/10.1111/cobi.12840 (2017).
Theodorou, P. et al. Urban areas as hotspots for bees and pollination but not a panacea for all insects. Nature communications, 11, 576; https://doi.org/10.1038/s41467-020-14496-6(2020).
Tew, N. E. et al. Quantifying nectar production by flowering plants in urban and rural landscapes. Journal of Ecology, 109, 1747-1757; https://doi.org/10.1111/1365-2745.13598 (2021).
Casanelles-Abella, J., & Moretti, M. Challenging the sustainability of urban beekeeping using evidence from Swiss cities. npj Urban Sustainability, 2, 3; https://doi.org/10.1038/s42949-021-00046-6 (2022).
Mahé, C., Jumarie, C., & Boily, M. The countryside or the city: Which environment is better for the honeybee? Environmental Research, 195, 110784. https://doi.org/10.1016/j.envres.2021.110784 (2021).
Statistics Canada. The bee-list: honey, here’s some data from across the Canadian pollination. (2022a).
Kevan, P. G., Greco, C. F., & Belaoussoff, S. Log-normality of biodiversity and abundance in diagnosis and measuring of ecosystemic health: pesticide stress on pollinators on blueberry heaths. Journal of Applied Ecology, 34, 1122-1136; https://doi.org/10.2307/2405226 (1997).
Rortais, A., Arnold, G., Halm, M. P., & Touffet-Briens, F. Modes of honeybees exposure to systemic insecticides: estimated amounts of contaminated pollen and nectar consumed by different categories of bees. Apidologie, 36, 71-83; https://doi.org/10.1051/apido:2004071 (2005).
Giannini, T. C. et al. Pollination services at risk: bee habitats will decrease owing to climate change in Brazil. Ecological Modelling, 244, 127-131; https://doi.org/10.1016/j.ecolmodel.2012.06.035 (2012).
Soroye, P., Newbold, T., & Kerr, J. Climate change contributes to widespread declines among bumble bees across continents. Science, 367, 685-688; https://doi.org/10.1126/science.aax8591 (2020).
Potts, S. G. et al. Global pollinator declines: trends, impacts and drivers. Trends in Ecology and Evolution, 25, 345-353; https://doi.org/10.1016/j.tree.2010.01.007 (2010).
Harrison, T., & Winfree, R. Urban drivers of plant‐pollinator interactions. Functional Ecology, 29, 879-888; https://doi.org/10.1111/1365-2435.12486 (2015).
Tonietto, R. K., & Larkin, D. J. Habitat restoration benefits wild bees: A meta‐analysis. Journal of Applied Ecology, 55, 582-590; https://doi.org/10.1111/1365-2664.13012 (2018).
Theodorou, P. et al. Pollination services enhanced with urbanization despite increasing pollinator parasitism. Proceedings of the Royal Society B. 283, 20160561; https://doi.org/10.1098/rspb.2016.0561 (2016).
Dai, D. Racial/ethnic and socioeconomic disparities in urban green space accessibility: Where to intervene?. Landscape and Urban Planning, 102, 234-244; https://doi.org/10.1016/j.landurbplan.2011.05.002 (2011).
Li, H., & Liu, Y. Neighborhood socioeconomic disadvantage and urban public green spaces availability: A localized modeling approach to inform land use policy. Land Use Policy, 57, 470-478; https://doi.org/10.1016/j.landusepol.2016.06.015 (2016).
Rosofsky, A., Levy, J. I., Zanobetti, A., Janulewicz, P., & Fabian, M. P. Temporal trends in air pollution exposure inequality in Massachusetts. Environmental research, 161, 76-86; https://doi.org/10.1016/j.envres.2017.10.028 (2018).
Jerrett, M. et al. A GIS–environmental justice analysis of particulate air pollution in Hamilton, Canada. Environment and Planning A, 33, 955-973; https://doi.org/10.1068/a33137 (2001).
Loram, A., Warren, P. H., & Gaston, K. J. Urban domestic gardens (XIV): the characteristics of gardens in five cities. Environmental Management, 42, 361-376; https://doi.org/10.1007/s00267-008-9097-3 (2008).
Scarrow, R. Gentrified gardens. Nature Plants, 8, 724-724; https://doi.org/10.1038/s41477-022-01212-z (2022).
Lowenstein, D. M., Matteson, K. C., Xiao, I., Silva, A. M., & Minor, E. S. Humans, bees, and pollination services in the city: the case of Chicago, IL (USA). Biodiversity and Conservation, 23, 2857-2874; https://doi.org/10.1007/s10531-014-0752-0 (2014).
Burr, A., Schaeg, N., Muñiz, P., Camilo, G. R., & Hall, D. M. Wild bees in the city: reimagining urban spaces for native bee health. Consilience, 16, 106-131; https://www.jstor.org/stable/26188777 (2016).
Samuelson, A.E., Schürch, R. & Leadbeater, E. Dancing bees evaluate central urban forage resources as superior to agricultural land. Journal of Applied Ecology, 59, 79-88; https://doi.org/10.1111/1365-2664.14011 (2022).
Garbuzov, M., Schürch, R. & Ratnieks, F.L. Eating locally: dance decoding demonstrates that urban honey bees in Brighton, UK, forage mainly in the surrounding urban area. Urban Ecosystems, 18, 411-418; http://dx.doi.org/10.1007/s11252-014-0403-y (2015).
Hagler, J. R., Mueller, S., Teuber, L. R., Machtley, S. A., & Van Deynze, A. Foraging range of honey bees, Apis mellifera, in alfalfa seed production fields. Journal of Insect Science, 11, 144; https://doi.org/10.1673/031.011.14401 (2011).
Hashim, H., Abd Latif, Z., & Adnan, N. A. Urban vegetation classification with NDVI threshold value method with very high resolution (VHR) Pleiades imagery. The International Archives of Photogrammetry, Remote Sensing and Spatial Information Sciences, 42, 237-240; https://doi.org/10.5194/isprs-archives-XLII-4-W16-237-2019 (2019).
Government of Ontario. Air Quality Ontario: Pollutant Concentrations. (2023).
Samuelson, A.E., Gill, R.J. & Leadbeater, E. Urbanisation is associated with reduced Nosema sp. infection, higher colony strength and higher richness of foraged pollen in honeybees. Apidologie51, 746–762; https://doi.org/10.1007/s13592-020-00758-1 (2020).
Goddard M.A., Dougill A.J., Benton T.G. Scaling up from gardens: biodiversity conservation in urban environments. Trends in Ecology and Evolution, 25, 90–98; https://doi.org/10.1016/j.tree.2009.07.016 (2010).
Vanderplanck, M. et al. Ozone pollution alters olfaction and behavior of pollinators. Antioxidants, 10, 636; https://doi.org/10.1016/j.tree.2022.08.004 (2021).
Démares, F., Gibert, L., Creusot, P., Lapeyre, B., & Proffit, M. Acute ozone exposure impairs detection of floral odor, learning, and memory of honey bees, through olfactory generalization. Science of the Total Environment, 827, 154342; https://doi.org/10.1016/j.scitotenv.2022.154342 (2022).
Environmental Protection Agency. Ground-level Ozone Basics. (2022).
Leonard, R. J., Pettit, T. J., Irga, P., McArthur, C., & Hochuli, D. F. Acute exposure to urban air pollution impairs olfactory learning and memory in honeybees. Ecotoxicology, 28, 1056-1062; https://doi.org/10.1007/s10646-019-02081-7 (2019).
City of Montreal. Using Pesticides. (2023).
Ficetola, G.F., Rubolini, D. Climate affects global patterns of COVID-19 early outbreak dynamics. MedRxiv, 1-24; https://doi.org/10.1101/2020.03.23.20040501 (2020).
Attia, Y. A. et al. COVID-19 pandemic: impacts on bees, beekeeping, and potential role of bee products as antiviral agents and immune enhancers. Environmental Science and Pollution Research, 29, 1-14; https://doi.org/10.1007/s11356-021-17643-8 (2022).
Scivicco, M. et al. Effects of Covid-19 pandemic lockdown and environmental pollution assessment in Campania region (Italy) through the analysis of heavy metals in honeybees. Environmental Pollution, 307, 119504; https://doi.org/10.1016/j.envpol.2022.119504 (2022).
Statistics Canada. Has the COVID-19 pandemic changed commuting patterns for good? (2022b).
Tonietto, R., Fant, J., Ascher, J., Ellis, K., & Larkin, D. A comparison of bee communities of Chicago green roofs, parks and prairies. Landscape and Urban Planning, 103, 102-108; https://doi.org/10.1016/j.landurbplan.2011.07.004 (2011).
MacIvor, J. S. Building height matters: nesting activity of bees and wasps on vegetated roofs. Israel Journal of Ecology & Evolution, 62, 88-9; https://doi.org/10.1080/15659801.2015.1052635 (2016).
Iwasaki, J. M., & Hogendoorn, K. How protection of honey bees can help and hinder bee conservation. Current Opinion in Insect Science, 46, 112-118; https://doi.org/10.1016/j.cois.2021.05.005 (2021).
Martinello, M. et al. Spring mortality in honey bees in northeastern Italy: detection of pesticides and viruses in dead honey bees and other matrices. Journal of Apicultural Research, 56, 239-254; https://doi.org/10.1080/00218839.2017.1304878 (2017).
Doublet, V., Labarussias, M., de Miranda, J.R., Moritz, R.F.A., & Paxton, R.J. Bees under stress: Sublethal doses of a neonicotinoid pesticide and pathogens interact to elevate honey bee mortality across the life cycle. Environmental Microbiology,17, 969–983; https://doi.org/10.1111/1462-2920.12426 (2014).
Couvillon, M.J. et al. Honey bee foraging distance depends on month and forage type. Apidologie, 46, 61-70; https://doi.org/10.1007/s13592-014-0302-5 (2015).
Couvillon, M.J., Schürch R. & Ratnieks, F.L.W. Waggle Dance Distances as Integrative Indicators of Seasonal Foraging Challenges. PLoS ONE, 9, e93495; https://doi.org/10.1371/journal.pone.0093495 (2014).
Hammer, M. S. et al. Global estimates and long-term trends of fine particulate matter concentrations (1998–2018). Environmental Science & Technology, 54, 7879-7890; https://doi.org/10.1021/acs.est.0c01764 (2020).
Environment and Climate Change Canada. Air Quality Research Division, Toronto,Canada. (2017).
Bossu, A., Marco, A., Manel, S., & Bertaudière-Montes, V. Effects of built landscape on taxonomic homogenization: Two case studies of private gardens in the French Mediterranean. Landscape and urban planning, 129, 12-21; https://doi.org/10.1016/j.landurbplan.2014.05.002 (2014).
Gerner, E. E., & Sargent, R. D. Local plant richness predicts bee abundance and diversity in a study of urban residential yards. Basic and Applied Ecology, 58, 64-73; https://doi.org/10.1016/j.baae.2021.11.004 (2022).
Statistics Canada. 2016 Census of the Population. (2016).
Sibson, R. A brief description of natural neighbour interpolation. In Interpreting Multivariate Data (eds. Barnett, V.), 21-36, (1981).
Moroń, D. et al. Abundance and diversity of wild bees along gradients of heavy metal pollution. Journal of Applied Ecology, 49, 118–125; https://doi.org/10.1111/j.1365-2664.2011.02079.x (2012).
Dooling, R. J., & Popper, A. N. Some lessons from the effects of highway noise on birds. Proceedings of Meetings on Acoustics, 27, 010004; https://doi.org/10.1121/2.0000244 (2016).
Communauté métropolitaine de Montréal. Données géoréférencées. (2012).
Ontario Ministry of Natural Resources and Forestry. Ontario Road Network (ORN) Segment with Address. (2018).
Fox, J., Weisberg, S. An R Companion to Applied Regression, (Sage, 2019).
Akinwande, M. O., Dikko, H. G., & Samson, A. Variance inflation factor: as a condition for the inclusion of suppressor variable (s) in regression analysis. Open journal of statistics, 5, 754; https://doi.org/10.4236/ojs.2015.57075 (2015).
Bivand R, Wong D. Comparing implementations of global and local indicators of spatial association. TEST, 27, 716–748; https://doi.org/10.1007/s11749-018-0599-x (2018).

No competing interests reported.

Determinants of honeybee hive survival within cities and its use to monitor urban biodiversity: a Canadian study

Status:

Version 1

Abstract

Figures

Introduction

Results

Discussion

Methods