https://orcid.org/0000-0003-3523-6649
Figures
Abstract
Improving our understanding of how climate influences honey bee parasites and pathogens is critical as weather patterns continue to shift under climate change. While the prevalence of diseases vary according to regional and seasonal patterns, the influence of specific climatic predictors has rarely been formally assessed. To address this gap, we analyzed how occurrence and intensity of three prominent honey bee disease agents (Varroa destructor ― hereon Varroa ― Melissococcus plutonius, and Vairimorpha spp.) varied according to regional, temporal, and climatic factors in honey bee colonies across five Canadian provinces that were sampled at three time points. We found strong regional effects for all disease agents, with consistently high Varroa intensity and infestation probabilities and high M. plutonius infection probabilities in British Columbia, and year-dependent regional patterns of Vairimorpha spp. spore counts. Increasing wind speed and precipitation were linked to lower Varroa infestation probabilities, whereas warmer temperatures were linked to higher infestation probabilities. Analysis of an independent dataset shows that these trends for Varroa are consistent within a similar date range, but temperature is the strongest climatic predictor of season-long patterns. Vairimorpha spp. intensity decreased over the course of the summer, with the lowest spore counts found at later dates when temperatures were warm. Vairimorpha spp. intensity increased with wind speed and precipitation, consistent with inclement weather limiting defecation flights. Probability of M. plutonius infection generally increased across the spring and summer, and was also positively associated with inclement weather. These data contribute to building a larger dataset of honey bee disease agent occurrence that is needed in order to predict how epidemiology may change in our future climate.
Citation: McAfee A, Alavi-Shoushtari N, Tran L, Labuschagne R, Cunningham M, Tsvetkov N, et al. (2024) Climatic predictors of prominent honey bee (Apis mellifera) disease agents: Varroa destructor, Melissococcus plutonius, and Vairimorpha spp. PLOS Clim 3(8): e0000485. https://doi.org/10.1371/journal.pclm.0000485
Editor: Kevin Tougeron, Universite Catholique de Louvain, BELGIUM
Received: May 21, 2024; Accepted: August 12, 2024; Published: August 29, 2024
This is an open access article, free of all copyright, and may be freely reproduced, distributed, transmitted, modified, built upon, or otherwise used by anyone for any lawful purpose. The work is made available under the Creative Commons CC0 public domain dedication.
Data Availability: All raw data underlying these analyses are available as supplementary material. 2020 and 2021 pathogen, parasite, regional, temporal, climate, and GPS location data are available in Supplementary Data 1 & 2, respectively. Colony locations are also depicted in Supplementary Files 1 & 2. The 2016 Varroa and climate data are reported in Supplementary Data 3.
Funding: This work was funded by Genome BC through the Genomic Innovation for Regenerative Agriculture, Food and Fisheries (GIRAFF) program to LJF, LT, MMG, CAB, and JD. Data analyzed in this work was derived from the BeeCSI project, which was funded and supported by the Ontario Genomics Institute (OGI-185) to AZ, Genome Canada (LSARP #16420) to AZ, LJF, PG, SEH, EG, RWC, MMG and SFP, the Ontario Research Fund (LSARP #16420) to AZ, Genome Quebec, and the Government of Canada through Agriculture and Agri-Food Canada (AAFC) Genomics Research and Development Initiative (GRDI) funding (AAFC J-002368) to MMG and SFP. LT and EMW received salaries in part paid from AAFC GDRI funds, and MC received a stipend as Research Affiliate from AAFC as well as funds derived from the GIRAFF program. Salaries for AM, JC, HH, SKF, IMC, and DB were derived in part from funds awarded by Genome Canada. Salaries for AM and NT were also derived in part from funds derived from the GIRAFF program. This work was also supported by the Canadian Honey Council and the Technology Transfer Program (of which DB is a member) of the Ontario Beekeepers’ Association. The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript.
Competing interests: The authors have declared that no competing interests exist.
Introduction
Infectious diseases and parasites are among the dominant factors affecting honey bee (Apis mellifera) colony health in commercial operations [1–3]. Several recent studies have investigated relationships between climatic variables and overwintering colony mortality [4–8]. but similarly large-scale association studies between weather patterns and specific diseases or pathogens are more rarely conducted, with the notable exception of Rowland et al. [9], who examined patterns of disease incidence associated with climatic variables in England and Wales. Weather patterns are expected to become less stable as the climate changes, with higher frequencies of extreme events along with a shifting baseline to warmer temperatures and region-specific shifts in precipitation [10–12]. A better understanding of relationships between honey bee disease agents and current climatic trends would be an asset for predicting how these agents might change in the vicissitudes of our future climate.
Several etiological agents are known to occur in seasonal cycles, including but not limited to Vairimorpha spp. (formerly Nosema spp. [13], the causative agent(s) of nosemosis), Melissococcus plutonius (the causative agent of European foulbrood disease, or EFB), and the mite Varroa destructor (hereon referred to as Varroa, the causative agent of varroosis) [14–21]. Such seasonal patterns may be driven by direct effects of weather on pathogen or parasite survival and reproduction, or indirect effects via population or demographic changes within the honey bee colony. As an example, high temperatures (>35°C) reduce Varroa survival and reproduction directly [22, 23]. whereas moderately increasing temperatures during the spring favor Varroa population growth indirectly through increased honey bee brood rearing (which Varroa requires for reproduction) [14]. Moreover, favorable conditions (warm temperatures with low wind and precipitation) increase dispersal opportunities between colonies for disease agents, but might also affect the risk of colonies manifesting diseases (such as EFB [24]) that are associated with poor food availability or changes in the number of nurse bees available to tend to brood. It is thus difficult to predict, a priori, how weather conditions may translate into changes in intensity or occurrence of disease agents.
Varroa is broadly considered the most economically damaging parasite of honey bees, with only some remote islands remaining Varroa-free [14], and its life cycle is closely tied to seasonal changes in honey bee brood production [14, 25]. The mites disperse to new colonies through contact between bees at forage sources as well as robbing and drift (see reviews [14, 25] for more information about Varroa biology and control). Varroa has successfully become established in all continents where honey bees are maintained, and only a handful of island territories remain Varroa-free [14, 26]. Many studies broadly highlight the impact of winter brood breaks on Varroa loads and differences in Varroa population growth in temperate versus tropical climates (reviewed in Rosenkranz et al. [25]). The studies that have investigated relationships with weather conditions have focused mainly on temperature [9, 22, 23, 27], with extreme heat being associated with reduced Varroa reproduction [22, 23] but otherwise a positive relationship between Varroa occurrence and temperature has been observed [9, 27]. One study conducted by Harris et al. [23], found that rainfall had no association with Varroa population growth rates, and a recent study by Rowland et al. [9] found a negative correlation between Varroa occurrence and precipitation and wind. Further studies in different regions would be helpful to clarify and substantiate these findings in different geographies.
Vairimorpha spp. are microsporidian endoparasites of the honey bee midgut and, in temperate climates, tend to infect bees with seasonal intensities that are normally highest in the spring and early summer, then decrease in late summer and early fall [18–21]. Two species of Vairimorpha are known to infect honey bees in Canada: V. apis and V. ceranae with different seasonal trends. Whereas V. apis exhibits a regular peak in average gut spore loads in the spring and often a secondary peak in in the fall [28], V. ceranae trends appear more variable across different countries [29–32]. Regional differences in V. ceranae and V. apis abundance has led to speculation over whether each species’ relative success is influenced by climate, but it is not clear if this is the case [17, 19]. Some [18, 33] but not all [34] field studies have found that higher temperatures correlate with lower spore loads, and some research suggests that geographic region (and, by extension, climate) is an important factor influencing Vairimorpha spp. epidemiology [35]. Indeed, in addition to warm temperatures inhibiting V. apis and V. ceranae proliferation in laboratory trials [36], inclement (cold, windy, or wet) weather limits opportunities for defecation flights that help clear Vairimorpha spp. infections [37, 38], but interactions among these variables have, to our knowledge, not yet been investigated.
EFB disease can cause severe damage to honey bee colonies by killing developing brood, thereby weakening the colony and reducing productivity [16]. M. plutonius is often detected in asymptomatic and apparently healthy colonies, especially those proximal to symptomatic colonies and apiaries [16, 24, 39]; therefore, dispersal from infected units is likely an important factor governing patterns of M. plutonius prevalence. EFB is thought to be an opportunistic disease that tends to manifest when larvae are under nutritional stress, such as when colonies are producing new brood very quickly in the spring or when pollen and nectar resources are low [16, 24]. However, larval food supply is clearly not the only factor involved, since larvae fed in excess may still develop the disease [40]. Interestingly, Milbrath [16] and Bailey [41] suggest that even favorable foraging conditions could lead to temporary nutritional stress of larvae during the spring if a strong nectar flow causes nurse bees to reallocate their labor to process honey. Rowland et al. [9] found that higher precipitation (associated with poor foraging conditions) was linked to higher EFB incidence in England and Wales, but no relationship was found with temperature nor wind speed.
Given how widespread these honey bee pathogens and parasites are and their apparent associations with weather conditions, we sought to investigate relationships between the disease agents and climatic variables while accounting for broad differences between regions and years. To this end, we used a dataset, in part previously published [42, 43], of Varroa infestation levels, Vairimorpha spp. spore intensities, and M. plutonius occurrence across five Canadian provinces, representing a wide range of climates (including Köppen climate types of oceanic in British Columbia, cold semi-arid in southern Alberta, subarctic in northern Alberta, and humid continental in Manitoba, Ontario, and Quebec), to identify associations with temperature, precipitation, and wind speed. While regional differences were most pronounced, climatic variables were also influential and largely reveal intuitive patterns. This dataset will help build the growing body of knowledge and data necessary to facilitate predictive modelling of disease agent dynamics under projected climate conditions.
Methods
Honey bee colonies and datasets
This study includes analyses of a primary dataset (Varroa, Vairimorpha, and M. plutonius data collected in 2020 and 2021) and a Varroa validation dataset (Varroa only, collected in 2016). Some, but not all, elements of the primary and validation datasets have been previously described in several publications (see details below).
Primary dataset
A subset of the primary dataset’s honey bee colonies and pathogen/parasite data have been previously described by French et al. [42] (240 colonies sampled in 2020 and 2021) along with additional colonies (a further 240 in both years) described here and partially reported in McAfee et al. [43] As described further below, samples for M. plutonius and Vairimorpha spp. testing were pooled ahead of analysis (4 colonies per composite sample), whereas samples taken for Varroa detection were not pooled. The total number of samples analyzed for M. plutonius, Vairimorpha spp., and Varroa were therefore N = 120, 120, and 480, respectively, though all were derived from the same 480 colonies.
As previously described [42], the colonies were located in five provinces (British Columbia (BC), Alberta (AB), Manitoba (MB), Ontario (ON), and Quebec (QC)), and eight different regions. Within these regions, colonies were derived from the same beekeeping operation; therefore, operational and regional differences are indistinguishable. Two regions/operations were represented within AB (near Grande Prairie to the north and Lethbridge to the south); three regions/operations were represented in QC (near Quebec City in 2020 and near Montreal and Lac St. Jean in 2021); and BC, MB, and ON were represented by one region/operation each.
The initial purpose of the experiment from which these data are derived was to test for impacts of colony proximity to different focal crops on various colony health indicators. In the present manuscript, our preliminary analysis showed that proximity to focal crops was not an influential factor; nevertheless, we include a description of the experimental design here for full transparency. Colonies were placed near or far from one of eight possible focal crops, which depended on region, and sampled at the start (immediately before or immediately after moving into pollination), middle (peak bloom), and end (immediately before or immediately after) of the pollination period. The timing and duration of the pollination period varied depending on commercial standards for each crop, but broadly fell between the months of April and August. In the case of corn, which is not pollinated by honey bees, colonies were sampled before sowing (April), during sowing (May), and at the end of the growing season (September). An equal number (N = 240) of additional colonies produced from the same stocks, according to the same standards, as those assigned to the focal crops but located far from the pollination yards (>1.5 km, with the exception of highbush blueberries which were > 1.3 km) are included in this dataset, bringing the final number of colonies to 480 across site types (near or far from crops) in both years, each subject to three sampling events over time. A subset of these data are also described in McAfee et al. [43] Please see French et al. [42] and McAfee et al. [43] for more detailed descriptions around apiary site selection. Colony locations in each province and at each time point are illustrated in S1 Fig and sampling dates for each crop and region are exactly as described by French et al. [42] (S3 Table therewithin), and are also contained within S1 and S2 Data. We were unfortunately unable to obtain information on surrounding honey bee colony densities for this dataset.
In each region, colonies were either supplied by collaborating beekeepers, owned by the relevant institution, or purchased from producers, but in all cases sampling and management during the experiment were conducted by the research teams. As previously described [42], colonies were headed by locally overwintered queens and colony sizes were standardized in deep Langstroth-style single- or double- box pollination units according to standards for commercial pollination for each focal crop. Although Varroa is one of the parasites of interest, which we subsequently measured, early season Varroa control is often necessary to ensure colony vitality throughout the season; therefore, in some cases colonies were treated with miticides before the experiment began if adult bee infestation levels exceeded the economic threshold (>1 mite per 100 bees) [44]. Specifically, in early spring, colonies in BC (2020; n = 64) were treated with Formic Pro and colonies in southern Alberta (2020 and 2021; n = 48 in both years) were treated with Apivar according to manufacturer instructions. Note that although Apivar efficacy has waned in some regions [45], initial mite abundances in southern Alberta were very low (<1%); therefore, this potential issue did not have a meaningful impact on our results. Mite treatments applied to colonies prior to being purchased from commercial suppliers (BC in 2021 [n = 80] and MB in 2020 and 2021 [n = 56 and 16, respectively]) are unknown. All other colonies in northern AB (n = 16 in both years), ON (n = 40 in 2020), and QC (n = 16 in 2020 and 80 in 2021) were under university or government research laboratory management for the full calendar year and miticides were not applied. These differences in treatment (and other operational/regional differences) were accounted for statistically by including region as an interactive term in the model and by including colony as a random intercept term (accounting for differences in baseline mite levels). No colonies were treated between study commencement and termination, and no colonies were treated with Fumagillin-B (a product for controlling Vairimorpha spp.) or oxytetracycline (an antibiotic for controlling EFB disease).
Varroa validation dataset
The Varroa validation dataset consisted of previously published data quantifying Varroa infestations in Canada in 2016 (described by Borba et al. [46]). We used these data to validate our findings relating Varroa to regional, temporal, and climate variables and explore how length of beekeeping season relates to Varroa occurrence. For our analysis, we only used Varroa infestation data from overwintered colonies belonging to, and managed by, beekeeper collaborators where their mite loads were not manipulated and colonies were not otherwise managed in a way that could affect Varroa populations (colonies were not split, requeened, or allowed to swarm) for the duration of the experiment. This includes N = 480 overwintered colonies located in BC, southern AB, southern ON, and southern QC. Colonies within BC were located in 11 yards and were owned by 11 operators in coastal BC (<100 km from the coast; 8 yards) and interior BC (Thompson/Okanagan and Kootenay/Boundary regions; 3 yards), which have notably different climates. Colonies in AB, ON, and QC were also distributed across multiple yards, all of which are shown in Borba et al. [46] Each colony was sampled at up to three time points during the active beekeeping season. Data were filtered to exclude colonies from which fewer than two mite wash samples were obtained between May and September, as well as colonies which beekeepers moved to unknown locations between mite sampling events.
Pathogen testing (primary dataset)
Varroa, M. plutonius, and Vairimorpha 2020 and 2021 sample analysis was performed exactly as previously described [42]. Briefly, for M. plutonius and Vairimorpha, samples were analyzed as pooled replicates (15 bees from 4 colonies each, yielding n = 5 replicates from each site type (near to or far from focal crops) for each crop in each year. The same four colonies were pooled at each time point, and pooled colonies were always located together in the same yard (and therefore experienced the same climatic conditions). M. plutonius analysis was performed by endpoint PCR (forward primer: CAG CTA GTC GGT TTG GTT CC; reverse primer: TTG GCT GTA GAT AGA ATT GAC AAT) and Vairimorpha spp. spore counts were determined by microscopy (hemocytometer counting) and expressed as number of spores per bee [42]. Samples for Varroa mite counts were not pooled across colonies, and mite counts per 100 bees were determined using the alcohol wash method [47], modified slightly as previously described [46]. Therefore, final sample sizes for M. plutonius and Vairimorpha data were N = 60 pooled samples per time point, per year (360 samples from 120 pooled four-colony units), whereas the sample size for Varroa was N = 240 colonies per time point, per year (1,440 samples from 480 colonies). Please see the previously published methods for specific details regarding the protocols used [42].
Varroa testing (validation dataset)
The Varroa abundances in colonies contributing to the validation dataset were determined by the alcohol wash method [47], as described in Borba et al. [46], and expressed as mites per 100 bees.
Climatic data extraction
Datasets from Environment and Climate Change Canada climate stations [48] were used to extract hourly and daily climate variables (mean, maximum, and minimum daily temperatures, average daytime and nighttime wind speeds, and total precipitation). The climate stations that were within 30 km of the colony locations were selected and the daily and hourly datasets for the corresponding time of the sampling events were downloaded. The daily climate records were used to extract the daily climate variables for the sampling dates and were averaged over three weeks (21 days) prior to the sampling date. Averaging climatic variables across time prior to sampling dates improves model fits [9], and three weeks prior (corresponding to one honey bee brood cycle) was the longest period for which we had precise knowledge of colony GPS locations. Because wind is predominantly expected to influence honey bees via changing their flight behavior, and since flight only occurs during the daylight hours, we used the hourly climate records to extract daytime and nighttime averages of wind speed for the sampling dates, which were also averaged over three weeks prior to the sampling dates. Daytime was defined as 5 am to 8 pm for all locations and dates, except for September sampling in Ontario, for which daytime was defined as 6 am to 7 pm to more closely match actual sunrise and sunset times late in the season. For the colonies with more than one climate station within a 30 km distance, the climate data were extracted for each climate station independently as described above, then averaged across all climate stations within the 30 km distance.
Prior to conducting statistical modelling, correlations among climate variables in the primary dataset were assessed by computing Pearson correlation coefficients (S2 Fig). Based on these results, we included only mean temperature, average daytime wind speed, and total precipitation in our statistical models (described under “Statistical analysis,” as these variables had correlation coefficients of |r| < 0.4 in 2020 and < 0.3 in 2021. We chose to use daytime wind exclusively, because wind is expected to only impact honey bees during the day when they have an opportunity to fly. For precipitation, we used daily totals, rather than daytime totals, because precipitation is also linked to nectar production in flowers and ambient humidity regardless of when it falls, and could therefore conceivably affect honey bee behavior, thermoregulation, or nutrition at any time.
Statistical analysis
All statistical analyses were performed in R (version 4.3.0) using R Studio (version 2023.09.1+494) [49]. The primary dataset (2020 and 2021) and Varroa validation dataset (2016) were analyzed separately. For the primary dataset, site type (whether colonies/replicates were located near or far from focal crops) was first considered as a covariate but in no case was it a significant predictor, so it was dropped from the models. For all models, appropriateness of fit was confirmed using tools within the DHARMa package [50]. Interaction terms were removed from the model if they were non-influential and if appropriateness of fit was retained after doing so.
The Varroa data in the primary dataset are proportional (mites per 100 bees) and zero-heavy; therefore, we analyzed the data in two ways: first, we modelled the non-zero subset of the data (153/480 samples) to describe how intensity of Varroa infestation changed in different regions over time, then, we modelled binomial Varroa detections (occurrence) separately. In no cases (for Varroa nor other pathogens) were zeroes imputed. The non-zero Varroa data were transformed by taking the natural logarithm and analyzed using a linear mixed model (lme4 package [51]) with year (two levels) as a fixed effect, calendar date (continuous) and general region/operation (eight levels) as interacting effects, and colony as a random intercept term. Next, Varroa occurrence (1 = detected, 0 = not detected) from sampling dates at similar points in the beekeeping season (all in July for 2020 and all in June for 2021, no repeated sampling, years analyzed separately) were extracted from the complete dataset and analyzed using a generalized linear model, specifying a binomial distribution, and with region/operation (five levels in each year) included as a fixed effect. The Ontario and Quebec-Montreal regions were excluded from this analysis as they did not have a Varroa sample taken at a comparable date to the other regions.
To determine relationships with regional, temporal and climatic variables in the primary dataset, Varroa occurrence from all time points and regions were analyzed using a generalized linear mixed model according to the following formula: Occurrence ~ Temperature*Wind+Precipitation + Date*Region + Year + (1|Colony). Here and in all cases (except for the Varroa validation analysis; see below), the three climatic variables were initially included as a three-way interaction term, but the model was simplified if this interaction was not significant. If a climate variable did not significantly predict occurrence as a main or interactive effect, and if retaining the variable in the model did not significantly improve residual or simulated residual distributions, the variable was dropped. M. plutonius occurrence was modelled using the same principles, according to the formula: Occurrence ~ Wind*Precipitation + Date*Region + Year + (1|Unit), where unit refers to the pooled sample group. For Vairimorpha, non-zero intensities (spore loads; 305/360 samples) were first log10 transformed, then modelled using a linear mixed model according to the formula: Intensity ~ Wind*Precipitation + Temp + Date*Region + Year + (1|Unit). Vairimorpha data from Quebec in 2020 were removed as there were too few data points to draw meaningful comparisons (< n = 3 per site type and time point).
To validate the results of our primary analysis of Varroa occurrence and climate data, we used the same linear model to describe Varroa occurrence data on the 2016 validation dataset (except that year was not included, as this dataset only spans one year). This dataset comprised Varroa occurrence previously published by Borba et al. [46] (see “honey bee colonies and datasets” section). The climate data were downloaded and averaged for each site in the same manner as described for the 2020 and 2021 data. In a second evaluation, we analyzed Varroa occurrence at the three different sampling events (early season, mid-season, and late season) separately, as the mid-season sampling dates have the greatest temporal overlap with the 2020 and 2021 data. For this analysis, we used a generalized linear model (no random effects were specified as there was no repeated sampling of colonies within each sampling event). Mean daily temperature data (spatially averaged across climate stations within a 30 km radius of each site) for full 2016 calendar year were also obtained for the corresponding GPS locations and used to estimate the beginning and end of the brood rearing season. While we do not know the actual dates that substantial brood rearing commenced and ceased, we estimated it to begin on the date at which the seven-day sliding average temperature was at or above 5°C (when brood rearing intensifies [52]) for at least three consecutive days, and ending on the date at which the seven-day sliding average temperature declined below 5°C for ten or more consecutive days. We chose these criteria on the rationale that it guards against making inferences based on spurious fluctuations. While imprecise, the approach is meant to illustrate broad patterns of seasonal timing in different regions.
Results
Analysis of the primary dataset shows that Varroa intensity generally increased over time, as expected, but varied significantly by region and year (Fig 1A; see Table 1 for statistical reporting). Notably, Varroa intensities in BC’s Lower Mainland were generally high even early in the season (April-May), with levels comparable to late-season (Aug-Sept) intensities in other provinces. Indeed, selecting comparable sampling dates across provinces (July sampling in 2020 and June sampling in 2021) shows that Varroa prevalence in BC is high with respect to other provinces, with an 88% and 50% positivity rate in 2020 and 2021, respectively (Fig 1B and 1C).
A) Varroa intensities (non-zero mites per 100 bees) varied significantly by year and a date-by-region interaction. Statistical parameters are described in Table 1. BC = British Columbia, AB (S) = Southern Alberta, AB (N) = Northern Alberta, MB = Manitoba, ON = Ontario, QC (QC) = Quebec, near Quebec City, QC (Mon) = Quebec, near Montreal, QC (LSJ) = Quebec, near Lac St. Jean. B & C) Region effects illustrated as occurrence rates at each site, restricted to comparable sampling dates (July dates in 2020 and June dates in 2021, no repeated sampling). Logistic regression analysis shows significant regional differences in occurrence frequencies in 2020 (C; χ2 = 82.6, df = 4, p < 0.001) and less pronounced, but still significant differences in 2021 (D; χ2 = 18.3, df = 4, p = 0.0011). Numbers inside circles indicate sample sizes.
We next modelled Varroa occurrence with regional, temporal (date, and year) and climatic (mean temperature, average wind speed, and total precipitation) predictors to determine how weather patterns influence occurrence probability. Date and temperature were, predictably, strongly correlated with one another (Pearson’s r = 0.9; S2 Fig). Nevertheless, temperature still positively predicted Varroa occurrence beyond the covarying effect of date. Interestingly, although Varroa occurrence was generally high in BC’s Lower Mainland, occurrence probability did not increase appreciably with time, as in most other regions (Fig 2A and 2B). We also found a significant interactive effect between temperature and wind, with higher wind speeds predicting lower Varroa occurrence probability, but this effect was moderated by high temperatures (S3 Fig). Precipitation was also negatively associated with Varroa occurrence, and the lowest occurrence probabilities were observed when both wind speeds and precipitation were high (Fig 2C). This effect was consistent across 2020 and 2021. See Table 2 for complete statistical reporting.
Statistical parameters fully reported in Table 2. A & B) Fitted Varroa occurrence probabilities (1 = present, 0 = absent) vary by significant year (χ2 = 6.5, df = 1, p = 0.001) and a date-by-region interaction (χ2 = 29.6, df = 7, p < 0.001), among other influential factors. C) Occurrence probabilities significantly decrease with higher wind speeds, but the effect is reduced when temperatures are warm (wind-by-temperature interaction: χ2 = 4.8, df = 1, p = 0.029; see also S2 Fig). Occurrence also significantly decreases with precipitation (χ2 = 9.2, df = 1, p = 0.0024). Histograms show densities of fitted probabilities.
To determine if our findings agreed with an independent dataset, we analyzed a validation dataset of Varroa occurrence acquired in 2016 from BC, AB, ON, and QC using the same model parameters. Unlike the 2020 and 2021 data, this dataset covers a longer time period, spanning dates from April to November (Fig 3A). Climatic variables in the primary and validation datasets had comparable average magnitudes, although the maximum precipitation observed in the primary dataset was notably higher than the validation dataset (primary dataset: mean temperature = 15.1°C (2.9–22.7°C), mean precipitation = 2.0 mm (0.1–19.6 mm), mean wind speed = 9.5 km/h (3.9–17.5 km/h); validation dataset: mean temperature = 15.6°C (6.7–23.0°C), mean precipitation = 2.4 mm (0.1–9.5 mm), mean wind speed = 11.6 km/h (3.4–21.3 km/h)). Early season (April-May) occurrence probabilities were again highest in BC (coastal regions including the Lower Mainland) and low in QC, which might reflect differences in brooding initiation in these areas, as temperatures in BC yards exceeded 5°C earlier than yards in all other regions (S4 Fig); however, wind was not associated with an overall decrease in occurrence (S5 Fig). Rather, temperature was the ultimate climatic predictor (associated with higher occurrence probability), again above and beyond a simple covariation between temperature and date (see Table 3 for complete statistical reporting). We hypothesized that this may be due to the dataset covering a wider range of temperatures across a longer season, therefore, we inspected early, mid, and late-season occurrence separately. Indeed, the mid-season occurrence–which overlap with the 2020 and 2021 sampling dates to the greatest degree–do display a negative relationship with wind, but no significant relationship with precipitation (Fig 3B), partially supporting our hypothesis.
A) Validation dataset of Varroa intensity and occurrence in BC, AB (S), ON, and QC during the 2016 beekeeping season (divided by early, mid-season, and late sampling events). Data were modelled according to the same formula as the 2020/2021 dataset, with full statistical results described in Table 3. The positive effect of temperature on Varroa occurrence probability was exacerbated, not mitigated, under windier conditions (temperature-by-wind interaction: χ2 = 16.3, df = 1, p < 0.001; see also S4 Fig). B) Examining each sampling time point separately, only mid-season Varroa occurrence (n = 480) is negatively associated with wind (χ2 = 7.0, df = 1, p = 0.0081), but again, the negative effect of wind is mitigated by warm temperatures (temperature-by-wind interaction: χ2 = 4.4, df = 1, p = 0.046). Mid-season sampling (June to mid-August) dates have the greatest overlap with 2020 and 2021 sampling dates. Model parameters and outcomes for mid-season data are fully reported in Table 4.
Next, we used the primary dataset to analyze relationships between Vairimorpha and the same regional, temporal and climatic variables as for Varroa. Spores were detected in 95% of replicates (114/120) for at least one time point; therefore, we used a linear mixed model with regional, temporal, and climatic variables as predictors of intensity and did not model Vairimorpha presence/absence. We again observed strong regional effects, but the patterns varied greatly from year to year (Fig 4A and 4B; see Table 5 for statistical reporting). For example, while patterns were similar in BC and southern AB across years, samples from northern AB and MB tended to have higher intensities in 2020 than 2021. While patterns over time within regions varied, overall spore intensities decreased as temperatures warmed and the season progressed (Fig 4C), but the effect of temperature was marginally not significant (Type II Wald χ2 test; χ2 = 3.09, df = 1, N = 480, p = 0.079) after accounting for the effect of date (Table 5). Wind speed and precipitation significantly interacted, with spore intensities increasing with higher precipitation, but only under windier conditions (Fig 4D).
Statistical parameters are fully reported in Table 5. Spore intensities vary significantly by region (χ2 = 38.5, df = 6, p < 0.001) across 2020 (A) and 2021 (B). Spore intensities generally decrease over time during the spring and summer, but this trend varies by region (date-by-region interaction: χ2 = 14.2, df = 6, p = 0.027). Histograms show spore intensity bins. C) The general decrease over time coincides with increased temperatures (r = 0.9, p < 0.001; see also S2 Fig). The black triangle indicates the economic threshold of 1 million spores/bee. D) Spore intensities significantly increase with precipitation, but only under windy conditions (precipitation-by-wind interaction: χ2 = 9.4, df = 1, p = 0.0020), illustrated in three equidensity wind speed bins. BC = British Columbia, AB (S) = Southern Alberta, AB (N) = Northern Alberta, MB = Manitoba, ON = Ontario, QC (Mon) = Quebec, near Montreal, QC (LSJ) = Quebec, near Lac St. Jean.
M. plutonius occurrence varied by region as well, but there was no significant effect of year (Table 6). Overall detection probability was high in BC and markedly increased over time between May and June, whereas zero detections were recorded in ON throughout the season (Fig 5A). Temperature was not a significant predictor after accounting for date, and was removed from the model; however, wind and precipitation significantly interacted, with increased M. plutonius occurrence probability under wetter, windier conditions as well as drier, less windy conditions, and lower occurrence probability under intermediate conditions (Fig 5B).
Occurrence was fitted with generalized linear mixed model with statistical parameters reported in Table 6. A) Occurrence varies significantly by region (χ2 = 25, df = 7, p < 0.001). B) Wind and precipitation significantly interact (χ2 = 6.7, df = 1, p = 0.0099), as illustrated by plotting fitted probabilities against precipitation in three equidensity wind bins, with M. plutonius occurrence decreasing with precipitation when wind speed is low, but increasing with precipitation when wind is medium or high. BC = British Columbia, AB (S) = Southern Alberta, AB (N) = Northern Alberta, MB = Manitoba, ON = Ontario, QC (QC) = Quebec, near Quebec City, QC (Mon) = Quebec, near Montreal, QC (LSJ) = Quebec, near Lac St. Jean.
Discussion
We confirm several trends reported or proposed elsewhere [9, 16, 38, 41], and add clarity to relationships that have been more inconsistent [18, 33, 34], which are discussed further below. By analyzing data from colonies across Canada, we are also able to comment on prevalence of disease agents in some of the major beekeeping regions in this country; most notably, that Varroa and M. plutonius are more prevalent in BC than other provinces, at least within our sampling breadth, which was concentrated on two operations in the Lower Mainland area near Vancouver. Although the sampling periods in our primary dataset were relatively short (typically spanning several weeks, and not the full beekeeping season), we still observe significant and interesting trends.
Climatic effects
We found that temperature, wind, and precipitation were all influential factors for predicting Varroa occurrence, even after accounting for regional/operation and temporal effects, with temperature having a positive effect while wind and precipitation had a negative effect. That temperature has a positive impact above and beyond the effect of date aligns with expectations, as warmer temperatures promote brood nest expansion and foraging, giving the mites more opportunities to reproduce and possibly disperse. The negative effect of precipitation and wind are also intuitive, as such conditions limit bee flight (reducing dispersal opportunities), as well as possibly brood production via reduced foraging (reducing mite population growth rates).
Interestingly, Rowland et al. [9] also found a positive effect of temperature and negative effects of wind and precipitation in an analysis of Varroa prevalence in the UK, which is consistent with our primary findings. However, our primary findings were only partially validated by our analysis of an independent Varroa dataset derived from similar regions in Canada [46]. There, we found that temperature was the dominant climatic driver and the negative effect of wind was only observed among samples taken mid-season (June to early August), which is the sample point overlapping the most with the data in our initial analysis. Temperature and wind metrics had similar average magnitudes and spanned similar ranges in both the primary and validation datasets, so we expected to observe similar effects. We suspect that the different results are largely due to the two datasets being acquired over different time scales: while the 2020 and 2021 data were acquired from samples typically taken 2–4 weeks apart during the late spring and summer (with the exception of Ontario in 2020), the 2016 validation dataset spanned a considerably longer portion of the beekeeping season. It is sensible that the climatic variable promoting reproduction (an exponential process), rather than dispersal, would be most influential in this temporal scale.
Some research suggests that Varroa reproduction is inhibited when ambient temperatures surpass 35°C [22, 23], and such conditions will likely become more common as the climate changes and the frequency of extreme heat events increases [53–55]. Hot temperatures can also be harmful to honey bees [7, 56, 57], but not always [58]. British Columbia and Alberta experienced a severe heat wave between June 25 and July 2, 2021 [59], which falls within our study period. Temperatures reached >40°C in some parts of the Lower Mainland during this time, but the event was not strongly reflected in our data due to a combination of there being lower temperatures at our specific apiary locations and our process of computing three-week averages of daily means. Unfortunately, our sampling did not continue for long enough after the heat dome to measure any potential effect on Varroa reproduction.
Reported relationships between Vairimorpha spore intensities and temperature have been somewhat inconsistent, and our data add some clarity to the matter. A number of studies highlight broad seasonal patterns (with spore intensities decreasing between spring and fall) [18, 19, 21, 32], and some describe a negative relationship with temperature which is likely due to increased flight days [18, 33], or possibly reduced ability to proliferate [36], though this is contradicted by Rangel et al. [34]. In our analysis, we found that temperature did not significantly predict a decline in spore intensities beyond the covarying effect of date, which supports the idea that flight days, and not temperature per se, is the limiting factor. Further support of this reasoning is the interaction we observed between precipitation and wind (Fig 4D). These variables limit defecation flight opportunities independent of temperature, and while neither was sufficient to predict spore intensities on their own, precipitation did promote higher spore intensities under windier conditions. This makes sense given that the average total precipitation observed was not extreme (typically not more than 5 mm) and was therefore unlikely to sufficiently limit flights on its own. The likely crux of defecation flights as the operational factor could also be why Rangel et al. found no effect of temperature in their study of Vairimorpha in feral colonies in Texas. Those samples were collected in July over the course of 20 years, when one would expect the temperatures to be more than sufficient to enable flight days regardless of year-to-year deviations.
M. plutonius is often described as an opportunistic pathogen that is triggered to cause EFB disease based on a combination of nutritional deficiencies [16, 24, 60], pathogen genotype [61, 62], and host genotype [63], among other factors. Given the possible connection with nutrition, it is conceivable that inopportune foraging conditions could promote disease and increase M. plutonius occurrence, but it has been suggested that strong nectar flows might actually also promote M. plutonius proliferation due to diversion of nursing care [16, 41]. While Wardell et al. [60] states that no obvious climatic conditions account for disease incidence [60], this is not surprising for such a multifactorial issue, and others have postulated that climate may indeed play a role in part due to the strong regional differences that have been observed [24, 61].
For M. plutonius, we found both strong regional differences and a relationship with wind and precipitation, but no relationship with temperature after accounting for sampling date (Fig 5; Table 6). The highest occurrence frequencies were observed under both favorable (low wind, low precipitation) and unfavorable (high wind, high precipitation) conditions, with lower occurrence in intermediate zones. This pattern, while at first puzzling, may actually reflect both predictions: increased foraging leads to a temporarily poorer brood care due to a shift from nursing to nectar-processing activities, while suboptimal foraging conditions also lead to poorer brood care via insufficient nutrition. It is also possible that favorable conditions lead to higher M. plutonius occurrence probability because this increases opportunities for dispersal, which is particularly relevant for this highly transmissible pathogen [16]. Rowland et al. previously reported that high precipitation, but not wind, was positively linked to higher EFB incidence [9], and our results may differ in part because we analyzed M. plutonius occurrence, which is distinct from clinical diagnosis of EFB disease. The two metrics are different but nevertheless connected, especially as only 100–200 M. plutonius bacteria are required to cause disease [40, 41]. We expect that our method of PCR detection, as opposed to disease detection, would be more sensitive to dispersal events in which the pathogen is transmitted but does not necessarily cause disease, which is consistent with our observation of higher occurrence under favorable climatic conditions. Future work may benefit from incorporating M. plutonius intensity, as measured by qPCR, in addition to occurrence in order to better understand climatic relationships.
Regional effects
We observed strong regional effects for all three disease agents, with distinctly high incidence of M. plutonius and Varroa in BC, and variable levels of Vairimorpha spp. from year to year. However, we note that in our experiment we are unable to distinguish between effects of region and effects of beekeeping operation. We and others have previously observed an association between EFB and highbush blueberry pollination [16, 60, 64], and there is a high density of highbush blueberry fields in British Columbia’s Lower Mainland; however, two controlled field trials (including one derived from these data) do not support the notion that engaging in blueberry pollination causes M. plutonius infection [43, 65]. Therefore, perceived associations may actually be due to the seasonal timing of blueberry pollination. EFB outbreaks have also been observed in parts of Europe, where authorities responded with strict control measures including hive destruction and exclusion zones around affected apiaries [66]. Especially given the high transmissibility of M. plutonius to neighboring colonies [16, 24, 39], strict control measures are essential for containment. BC has no provincial mandate for destroying or otherwise controlling affected colonies and apiaries, whereas in Ontario, where we observed zero M. plutonius detections, EFB is a reportable disease with enforceable movement restrictions and sanitation procedures.
British Columbia, at least in our sampling region of the Lower Mainland, also appears to suffer from high incidence of Varroa mites, particularly early in the season. This is despite colonies being treated for Varroa (in early April) before the beginning of the study. Although we did not formally test this, we speculate that this may be in part due to the region having notably mild winters, which lead to shorter brood breaks and earlier recommencement of brood rearing. These conditions mean that beekeepers have a shorter broodless window during which to apply a miticide treatment, and remaining mites are able to begin reproducing earlier. However, we note that the 2016 dataset contained yards in BC’s interior (the Thompson/Okanagan and Kootenay/Boundary regions), as well as the Lower Mainland and coastal regions, but the patterns observed there do not entirely agree with this idea. Based on average daily temperatures in the interior and Lower Mainland/coastal apiaries for 2016, we estimated that the spring brooding date in the interior was later, similar to the regions in other provinces. However, we still observed relatively high early-season Varroa incidence in the interior, with fitted probabilities of ~0.5 (higher than all other regions). Therefore, factors yet unaccounted for are clearly at play, such as possibly beekeepers’ winter and early spring mite management procedures, and the timing of these in relation to when the experiment began.
A factor which we were unable to account for in our study is the density of honey bee colonies in the surrounding area. This is unfortunate, as some of our data suggest that climatic factors may impact pathogen and parasite prevalence via dispersal opportunities, and dispersal opportunities are clearly linked to surrounding colony densities. Future work in this realm would benefit from experimentally manipulating colony densities to determine if density and climatic factors interact as expected.
Conclusion
We present an analysis of common honey bee disease agents and how they vary according to regional, temporal, and climatic factors. Our results largely support patterns expected based on existing data and knowledge of how these disease agents reproduce and disperse. Nevertheless, confirmation of these trends is important on a nation-wide scale and across many climatic and ecotypes. Our analysis also provides additional insight into more ambiguous associations, particularly for how temperature affects Vairimorpha spore intensities and relationships between weather and M. plutonius occurrence. In most of our models, we observed a large degree of unexplained variation, suggesting that other factors, such as bee genetics, coinfections, or landscape factors may also influence pathogen and parasite dynamics, which were not investigated here. We hope that this and future studies will inform meaningful global predictive models of disease agents under future climate projections.
Supporting information
S1 Data. Excel file containing 2020 Varroa, M. plutonius, and Vairimorpha spp. data, along with covariate information (primary dataset).
https://doi.org/10.1371/journal.pclm.0000485.s001
(XLSX)
S2 Data. Excel file containing 2021 Varroa, M. plutonius, and Vairimorpha spp. data, along with covariate information (primary dataset).
https://doi.org/10.1371/journal.pclm.0000485.s002
(XLSX)
S3 Data. Excel file containing the 2016 Varroa validation data, full-season temperature data, along with colony covariate information (Varroa validation dataset).
https://doi.org/10.1371/journal.pclm.0000485.s003
(XLSX)
S1 File. R markdown (Rmd) file containing the R code used to statistically analyze the primary dataset Varroa analysis and generate Figs.
https://doi.org/10.1371/journal.pclm.0000485.s004
(RMD)
S2 File. R markdown (Rmd) file containing the R code used to statistically analyze the primary dataset M. plutonius and Vairomorpha spp. analysis and generate Figs.
https://doi.org/10.1371/journal.pclm.0000485.s005
(RMD)
S3 File. R markdown (Rmd) file containing the R code used to statistically analyze the validation dataset Varroa analysis and generate Figs.
https://doi.org/10.1371/journal.pclm.0000485.s006
(RMD)
S1 Fig. Pdf file containing S1 Fig (map of colony locations).
https://doi.org/10.1371/journal.pclm.0000485.s007
(PDF)
S2 Fig. Pdf file containing S2 Fig (Pearson correlation matrix of climate variables).
https://doi.org/10.1371/journal.pclm.0000485.s008
(PDF)
S3 Fig. Pdf file containing S3 Fig (effect of wind speed and temperature on Varroa occurrence).
https://doi.org/10.1371/journal.pclm.0000485.s009
(PDF)
S4 Fig. Pdf file containing S4 Fig (Varroa validation dataset–location effects).
https://doi.org/10.1371/journal.pclm.0000485.s010
(PDF)
S5 Fig. Pdf file containing S5 Fig (temperature, not wind, drives Varroa occurrence in long-term data).
https://doi.org/10.1371/journal.pclm.0000485.s011
(PDF)
Acknowledgments
We would like to acknowledge A. Achkanian, R. Bahreini, S. Bailey, V. Beger, C. Bryant, A. Chapman, M. Chihata, C. Currie, K. Dave, G.L. Ganesan, P. Grubiak, D. Holder, A. Ibrahim, M.A. Imrit, J. Janisse, J. Kearns, K. Lynch, D. Micholson, M. Munro, I. Ovinge, M. Peirson, K. Peters, Z. Rempel, R. Thygesen, D. Tran, A. Travas, and B. Vinson for their supportive efforts. We are also grateful to all members of the BeeCSI Consortium (https://beecsi.ca/beecsi-consortium) who participated in this project.
References
- 1. Smith KM, Loh EH, Rostal MK, Zambrana-Torrelio CM, Mendiola L, Daszak P. Pathogens, pests, and economics: Drivers of honey bee colony declines and losses. Ecohealth. 2013;10(4):434–45. Epub 2014/02/05. pmid:24496582.
- 2. vanEngelsdorp D, Meixner MD. A historical review of managed honey bee populations in Europe and the United States and the factors that may affect them. Journal of Invertebrate Pathology. 2010;103:80–95. pmid:19909973
- 3. Steinhauer N, Kulhanek K, Antúnez K, Human H, Chantawannakul P, Chauzat M-P. Drivers of colony losses. Current Opinion in Insect Science. 2018;26:142–8. pmid:29764654
- 4. Switanek M, Crailsheim K, Truhetz H, Brodschneider R. Modelling seasonal effects of temperature and precipitation on honey bee winter mortality in a temperate climate. Science of the Total Environment. 2017;579:1581–7. pmid:27916302
- 5. Beyer M, Junk J, Eickermann M, Clermont A, Kraus F, Georges C, et al. Winter honey bee colony losses, Varroa destructor control strategies, and the role of weather conditions: Results from a survey among beekeepers. Research in Veterinary Science. 2018;118:52–60.
- 6. Calovi M, Grozinger CM, Miller DA, Goslee SC. Summer weather conditions influence winter survival of honey bees (Apis mellifera) in the northeastern United States. Scientific Reports. 2021;11(1):1553.
- 7. Insolia L, Molinari R, Rogers SR, Williams GR, Chiaromonte F, Calovi M. Honey bee colony loss linked to parasites, pesticides and extreme weather across the United States. Scientific Reports. 2022;12(1):20787. pmid:36456591
- 8. Becsi B, Formayer H, Brodschneider R. A biophysical approach to assess weather impacts on honey bee colony winter mortality. Royal Society Open Science. 2021;8(9):210618. pmid:34631120
- 9. Rowland BW, Rushton SP, Shirley MDF, Brown MA, Budge GE. Identifying the climatic drivers of honey bee disease in England and Wales. Scientific Reports. 2021;11(1):21953. Epub 20211109. pmid:34754028; PubMed Central PMCID: PMC8578631.
- 10. Salinger MJ. Climate variability and change: Past, present and future–an overview. Climatic Change. 2005;70(1–2):9–29.
- 11. Stott P. How climate change affects extreme weather events. Science. 2016;352(6293):1517–8.
- 12. Thornton PK, Ericksen PJ, Herrero M, Challinor AJ. Climate variability and vulnerability to climate change: A review. Global Change Biology. 2014;20(11):3313–28. pmid:24668802
- 13. Tokarev YS, Huang W-F, Solter LF, Malysh JM, Becnel JJ, Vossbrinck CR. A formal redefinition of the genera Nosema and Vairimorpha (Microsporidia: Nosematidae) and reassignment of species based on molecular phylogenetics. Journal of Invertebrate Pathology. 2020;169:107279. pmid:31738888
- 14. Traynor KS, Mondet F, de Miranda JR, Techer M, Kowallik V, Oddie MAY, et al. Varroa destructor: A complex parasite, crippling honey bees worldwide. Trends in Parasitology. 2020;36(7):592–606. Epub 2020/05/23. pmid:32456963.
- 15.
Bailey L, Ball B. Honey Bee Pathology, 2nd Edition. London: London, UK. Academic Press, Ltd; 1991.
- 16.
Milbrath M. Honey bee bacterial diseases. Kane TR, Faux CM, editors. New York: John Wiley & Sons, Inc; 2021. 277–93 p.
- 17. Gisder S, Hedtke K, ckel N, Frielitz M-C, Linde A, Genersch E. Five-year cohort study of Nosema spp. in Germany: does climate shape virulence and assertiveness of Nosema ceranae? Applied and Environmental Microbiology. 2010;76(9):3032–8. pmid:20228103
- 18. Retschnig G, Williams GR, Schneeberger A, Neumann P. Cold ambient temperature promotes Nosema spp. intensity in honey bees (Apis mellifera). Insects. 2017;8(1):20. pmid:28208761
- 19. Punko RN, Currie RW, Nasr ME, Hoover SE. Epidemiology of Nosema spp. and the effect of indoor and outdoor wintering on honey bee colony population and survival in the Canadian Prairies. PLoS One. 2021;16(10):e0258801. pmid:34695141
- 20. Guzman-Novoa E, Eccles L, Calvete Y, Mcgowan J, Kelly PG, Correa-Benitez A. Varroa destructor is the main culprit for the death and reduced populations of overwintered honey bee (Apis mellifera) colonies in Ontario, Canada. Apidologie. 2010;41(4):443–50. WOS:000280290600004.
- 21. Emsen B, De la Mora A, Lacey B, Eccles L, Kelly PG, Medina-Flores CA, et al. Seasonality of Nosema ceranae infections and their relationship with honey bee populations, food stores, and survivorship in a North American region. Veterinary Sciences. 2020;7(3):131. pmid:32911814
- 22. Le Conte Y, Arnold G, Desenfant P. Influence of brood temperature and hygrometry variations on the development of the honey bee ectoparasite Varroa jacobsoni (Mesostigmata: Varroidae). Environmental Entomology. 1990;19(6):1780–5.
- 23. Harris J, Harbo J, Villa J, Danka R. Variable population growth of Varroa destructor (Mesostigmata: Varroidae) in colonies of honey bees (Hymenoptera: Apidae) during a 10-year period. Environmental Entomology. 2003;32(6):1305–12.
- 24. Forsgren E. European foulbrood in honey bees. Journal of Invertebrate Pathology. 2010;103:S5–S9. pmid:20105559
- 25. Rosenkranz P, Aumeier P, Ziegelmann B. Biology and control of Varroa destructor. Journal of Invertebrate Pathology. 2010;103:S96–S119.
- 26. Owen R, Stevenson M, Scheerlinck J-P. Varroa destructor detection in non-endemic areas. Apidologie. 2021;52:900–14.
- 27. Smoliński S, Langowska A, Glazaczow A. Raised seasonal temperatures reinforce autumn Varroa destructor infestation in honey bee colonies. Scientific Reports. 2021;11(1):22256. pmid:34782664
- 28. Bailey L. The epidemiology and control of Nosema disease of the honey‐bee. Annals of Applied Biology. 1955;43(3):379–89.
- 29. Martín‐Hernández R, Botías C, Bailón EG, Martínez‐Salvador A, Prieto L, Meana A, et al. Microsporidia infecting Apis mellifera: coexistence or competition. Is Nosema ceranae replacing Nosema apis? Environmental Microbiology. 2012;14(8):2127–38. pmid:22176602
- 30. Higes M, Martín‐Hernández R, Botías C, Bailón EG, González‐Porto AV, Barrios L, et al. How natural infection by Nosema ceranae causes honeybee colony collapse. Environmental Microbiology. 2008;10(10):2659–69.
- 31. Martín-Hernández R, Meana A, Prieto L, Salvador AM, Garrido-Bailón E, Higes M. Outcome of colonization of Apis mellifera by Nosema ceranae. Applied and Environmental Microbiology. 2007;73(20):6331–8.
- 32. Copley T, Chen H, Giovenazzo P, Houle E, Jabaji S. Prevalence and seasonality of Nosema species in Québec honey bees. The Canadian Entomologist. 2012;144(4):577–88.
- 33. Chen Y-W, Chung W-P, Wang C-H, Solter LF, Huang W-F. Nosema ceranae infection intensity highly correlates with temperature. Journal of Invertebrate Pathology. 2012;111(3):264–7.
- 34. Rangel J, Baum K, Rubink WL, Coulson RN, Johnston JS, Traver BE. Prevalence of Nosema species in a feral honey bee population: a 20-year survey. Apidologie. 2016;47:561–71.
- 35. Pacini A, Molineri A, Antúnez K, Cagnolo NB, Merke J, Orellano E, et al. Environmental conditions and beekeeping practices associated with Nosema ceranae presence in Argentina. Apidologie. 2021;52:400–17.
- 36. Martín-Hernández R, Meana A, García-Palencia P, Marín P, Botías C, Garrido-Bailón E, et al. Effect of temperature on the biotic potential of honeybee microsporidia. Applied and Environmental microbiology. 2009;75(8):2554–7. pmid:19233948
- 37. Moeller F. Effects of emerging bees and of winter flights on nosema disease in honeybee colonies. Journal of Apicultural Research. 1972;11(2):117–20.
- 38. Meixner MD, Francis RM, Gajda A, Kryger P, Andonov S, Uzunov A, et al. Occurrence of parasites and pathogens in honey bee colonies used in a European genotype-environment interactions experiment. Journal of Apicultural Research. 2014;53(2):215–29.
- 39. Belloy L, Imdorf A, Fries I, Forsgren E, Berthoud H, Kuhn R, et al. Spatial distribution of Melissococcus plutonius in adult honey bees collected from apiaries and colonies with and without symptoms of European foulbrood. Apidologie. 2007;38(2):136–40.
- 40. McKee BA, David Goodman R, Alan Hornitzky M. The transmission of European foulbrood (Melissococcus plutonius) to artificially reared honey bee larvae (Apis mellifera). Journal of Apicultural Research. 2004;43(3):93–100.
- 41. Bailey L. The epizootiology of European foulbrood of the larval honey bee, Apis mellifera Linnaeus. Journal of Insect Pathology. 1960;2(2):67–83.
- 42. French SK, Pepinelli M, Conflitti IM, Jamieson A, Higo H, Common J, et al. Honey bee stressor networks are complex and dependent on crop and region. Current Biology. 2024;34:1893–903. pmid:38636513
- 43. McAfee A, French SK, Wizenberg SB, Newburn LR, Tsvetkov N, Higo H, et al. Higher prevalence of sacbrood virus in Apis mellifera (Hymenoptera: Apidae) colonies after pollinating highbush blueberries. Journal of Economic Entomology. 2024:toae119. pmid:38877967
- 44.
Currie RW. Economic threshold for Varroa on the Canadian Prairies. Canadian Association of Professional Apiculturists (CAPA) http://capabeesorg/content/uploads/2013/02/varroathresholdpdf. 2008.
- 45. Rinkevich FD. Detection of amitraz resistance and reduced treatment efficacy in the Varroa Mite, Varroa destructor, within commercial beekeeping operations. PLoS One. 2020;15(1):e0227264. pmid:31951619
- 46. Borba RS, Hoover SE, Currie RW, Giovenazzo P, Guarna MM, Foster LJ, et al. Phenomic analysis of the honey bee pathogen-web and its dynamics on colony productivity, health and social immunity behaviors. PLoS One. 2022;17(1):e0263273. pmid:35100308
- 47. De Jong D, Roma DDA, Goncalves L. A comparative analysis of shaking solutions for the detection of Varroa jacobsoni on adult honeybees. Apidologie. 1982;13(3):297–306.
- 48.
Government of Canada Environment and Natural Resources. Past weather and climate: Historical data: https://climate.weather.gc.ca/historical_data/search_historic_data_e.html [Accessed January 2024].
- 49.
R Core Team. R: A language and environment for statistical computing: R Foundation for Statistical Computing, Vienna, Austria; 2023.
- 50.
Hartig F. DHARMa: Residual diagnostics for hierarchical (multi-level/mixed) regression models. R package version 0.4.6. https://CRAN.R-project.org/package=DHARMa2022.
- 51. Bates D, Maechler M, Bolker B, Walker S. Fitting linear mixed-effects models esing {lme4}. Journal of Statistical Software. 2015;67(1):1–48.
- 52. Seeley TD, Visscher PK. Survival of honeybees in cold climates: the critical timing of colony growth and reproduction. Ecological Entomology. 1985;10(1):81–8.
- 53. Perkins-Kirkpatrick SE, Lewis SC. Increasing trends in regional heatwaves. Nature Communications. 2020;11(1):3357. Epub 20200703. pmid:32620857; PubMed Central PMCID: PMC7334217.
- 54. Meehl GA, Tebaldi C. More intense, more frequent, and longer lasting heat waves in the 21st century. Science. 2004;305(5686):994–7. pmid:15310900.
- 55. Luber G, McGeehin M. Climate change and extreme heat events. American Journal of Preventative Medicine. 2008;35(5):429–35. pmid:18929969.
- 56. McAfee A, Chapman A, Higo H, Underwood R, Milone J, Foster LJ, et al. Vulnerability of honey bee queens to heat-induced loss of fertility. Nature Sustainability. 2020;3:367–76.
- 57. Bach DM, Holzman MA, Wague F, Miranda JL, Lopatkin AJ, Mansfield JH, et al. Thermal stress induces tissue damage and a broad shift in regenerative signaling pathways in the honey bee digestive tract. Journal of Experimental Biology. 2021;224(18):jeb242262. pmid:34477881
- 58. Bordier C, Dechatre H, Suchail S, Peruzzi M, Soubeyrand S, Pioz M, et al. Colony adaptive response to simulated heat waves and consequences at the individual level in honeybees (Apis mellifera). Scientific Reports. 2017;7(1):3760. Epub 2017/06/19. pmid:28630407; PubMed Central PMCID: PMC5476575.
- 59. Thompson V, Kennedy-Asser AT, Vosper E, Lo YE, Huntingford C, Andrews O, et al. The 2021 western North America heat wave among the most extreme events ever recorded globally. Science Advances. 2022;8(18):eabm6860. pmid:35507648
- 60. Wardell G. European foulbrood: Association with Michigan blueberry pollination, and control: Michigan State University; 1982.
- 61. Budge GE, Shirley MD, Jones B, Quill E, Tomkies V, Feil EJ, et al. Molecular epidemiology and population structure of the honey bee brood pathogen Melissococcus plutonius. The ISME Journal. 2014;8(8):1588–97.
- 62. Grossar D, Kilchenmann V, Forsgren E, Charrière J-D, Gauthier L, Chapuisat M, et al. Putative determinants of virulence in Melissococcus plutonius, the bacterial agent causing European foulbrood in honey bees. Virulence. 2020;11(1):554–67. pmid:32456539
- 63. Ameline C, Beaurepaire A, Ory F, de La Harpe M, Dainat B, Dietemann V. Differential resistance across paternal genotypes of honey bee brood to the pathogenic bacterium Melissococcus plutonius. Journal of Applied Entomology. 2023;147(1):85–93.
- 64. Grant KJ, DeVetter L, Melathopoulos A. Honey bee (Apis mellifera) colony strength and its effects on pollination and yield in highbush blueberries (Vaccinium corymbosum). PeerJ. 2021;9:e11634. pmid:34395063
- 65. Fowler PD, Schroeder DC, Kevill JL, Milbrath MO. No impact of hygienic behavior and viral coinfection on the development of European foulbrood in honey bee (Apis mellifera) colonies during blueberry pollination in Michigan. Journal of Insect Science. 2023;23(6):21. pmid:38055945
- 66. Grossar D, Haynes E, Budge GE, Parejo M, Gauthier L, Charriere J-D, et al. Population genetic diversity and dynamics of the honey bee brood pathogen Melissococcus plutonius in a region with high prevalence. Journal of Invertebrate Pathology. 2023;196:107867.