Beekeepers regularly employ management practices to mitigate losses during the winter, often considered the most difficult time during a colony life cycle. Management recommendations involving covering or wrapping hives in insulation during winter have a long history; over 100 years ago, most recommendations for overwintering in cold climates involved heavy insulation wraps or moving hives indoors. These recommendations began to change in the mid-20^th century, but hive covers are still considered useful and are described in contemporary beekeeping manuals and cooperative extension materials. However, most of the data supporting their use is published primarily in non-peer reviewed trade journals and was collected >40 years ago. In this time, the beekeeping environment has changed substantially, with new pressures from pathogens, agrochemicals, and land use changes. Here, we provide an update to the historical literature, reporting a randomized experiment testing the effectiveness of a common honey bee hive cover system across eight apiaries in central Illinois, USA, a temperate region dominated by conventional annual agriculture. We found that, when other recommended overwintering preparations are performed, covered colonies consumed less food stores and survived better than uncovered controls (22.5% higher survival). This study highlights the value of hive covers, even in an area not subject to extremely cold winter conditions, and these data can aid the production of evidence-based extension recommendations for beekeepers.

Bee source and study sites: For this study, we used 43 full-sized honey bee colonies kept in standard 10-frame Langstroth hive bodies managed by the University of Illinois Bee Research Facility in Urbana, Illinois, USA (located at approximately 40°N) during 2020. Of these 22 were in the covered treatment group and 21 in the uncovered. All colonies were headed by a commercially sourced Carniolan (Apis mellifera carnica) or Italian (A. m. ligustica) honey bee queen or her direct descendant. Hives were kept across eight apiaries within Champaign County, with each apiary housing an average of 5.5 (±0.85 SEM) hives.

 Pre-overwintering management: On a monthly basis, we performed routine alcohol washes of approximately 300 nurse aged bees to monitor the population of the parasitic mite Varroa destructor within all colonies. All the source colonies in this study were treated for Varroa mites in early August, before equalization, using amitraz strips, Apivar (Mann Lake ltd.), per label instructions In October, mite levels varied significantly across the apiaries and were higher than the suggested threshold of 1% (mean 10.54 mites/300 bees (3.5% infestation), therefore, we treated all colonies using oxalic vaporization (OxaVap ProVap 110) per label instructions on October 12th. Mite levels were rechecked on November 9th and found to be reduced below threshold (mean of 2.2 mites/300 bees (0.7% infestation). At this time, there were no significant differences in mite load between colonies from the different, randomly selected, covered versus control treatment groups. A final oxalic acid vaporization treatment was performed on all colonies on January 6th. In all cases of mite treatments, every colony was treated identically, regardless of its mite load. On October 5th – 8th, all colonies were condensed into two 10-frame deep hive bodies and a single 10 frame medium honey super filled with capped honey from the summer 2020. At that same time, all colonies then received 1 gallon (3.79 liters) of liquid honey derived from our beekeeping operation, then two weeks later (October 23rd) received 1 gallon (3.79 liters) of 2:1 sucrose solution; both were delivered in 1 gallon division board feeders (Dadant and Sons. Inc., Hamilton IL). To ensure each colony possessed the necessary food stores for a successful overwinter, using a threshold of 30 kg of stored honey (per recommendations from (Doke et al 2019)), we measured the pre-overwintering weight on November 9th. All colonies were significantly heavier than this threshold (T7=15.42, p<0.0001), with an average weight by apiary of 63.62kg (±1.39kg SEM). There were significant differences in average colony mass between some apiaries (S2 Fig; F7,34=2.56, p=0.03), but there were no differences between the colonies that would later form the covered and control treatment groups (T14=1.40, p=0.18).

Wrapping and insulation: On November 12th, after colonies were condensed and supplementally fed, we randomly selected half of the focal colonies at each apiary to be placed into each treatment group (covered or control). Thus, at each apiary, we balanced the number of covered and control colonies when possible, resulting in a total of 22 covered and 21 uncovered controls across sites (S2 Table). Hive covers consisted of 4mm thick black corrugated polypropylene plastic sheets (Packaging Corporation of America in Conrad, IA) that were formed into rectangular prisms to slide over the top of the entire hive (Fig 2A), like those sold by Carters Honey Farms (https://www.cartershoneybees.com/product-page/overwinter-bee-hive-protective-cover). We placed 1.5-inch (3.81cm) foam insulation board (Owens Corning Foamular 250, R-7.5) on top of the inner covers of covered hives and top entrance holes were cut into each of the wraps to reduce condensation build up inside hives. Entrance reducers and mouse guards were placed onto all colonies (covered and control).

Mid-winter supplementation: To investigate whether colonies with covers consume a varied amount of mid-winter supplementation compared to uncovered control colonies, we placed a sugar cake patty that consisted of 7lbs. (3.18kg) dry granulated white sugar mixed with 1 cup (236.59 milliliters) of water in every colony. Sugar cakes were placed in the colonies on February 2nd when the external temperature was greater than 45°F (7.25°C) to reduce the risk of chilling bees within the colonies. Sugar cakes were placed in the colonies using a 1-inch (2.54cm) shim lined with 27 gauge 1/8th inch (0.32cm) wire mesh hardware cloth and were placed at the top of the colony above the honey super. The sugar cake is very solid, and little or no sugar passively falls into the hive from it. To place the shim on colonies we removed the inner cover board. To account for the change in mass of equipment added and removed during the mid-winter supplementation all subsequent colony mass checks (see mass monitoring below) were adjusted as follows, ((current mass – mass of removed cover board) + mass of shim). This adjustment did not account for the 7lbs of added sugar cake as that weight would be variable across colonies as they consume the feed; however, this addition was consistent across all colonies when added on February 2nd. At the end of the wintering period when the first spring inspection was conducted (March 30th) the remaining sugar cake was removed from each colony after mass checks were recorded, brought back to the lab, dehydrated in a laboratory drying oven for 12 hours, and the mass of the remaining granulated sugar quantified.

Temperature monitoring: To monitor the internal temperature of colonies, we placed a single thermocron iButton temperature meter (ibuttonlink.com) in each colony on November 9th. Temperature meters were placed between a strip of clear tape and inserted between two frames at the center of the uppermost deep sized hive body directly below the honey super. In addition to experimental colonies, we placed an iButton in one sentinel hive at three apiaries (HII, PT, and PF; total of 3 sentinel hives) to capture the variation in the ambient temperature compared to colony thermoregulatory temperatures. Sentinel hives consisted of a single Langstroth deep filled with 10 frames of drawn comb but empty of bees. The iButtons were set to record temperature within colonies once every 256 minutes from Nov. 9th through March 30th. Because temperatures varied across the course of a day, we calculated the average temperature of a colony per calendar week to use in our analysis.

Survivorship monitoring: To track colony survivorship throughout the winter, we used a medical stethoscope on the side of the hive body to listen for an audible buzz within. If no buzzing was heard, we used one sharp knock to attempt hearing another buzz. If no sounds were detected within a colony, that colony was considered dead for that timepoint. If, at the next timepoint that colony audibly buzzed then it would be considered alive at all previous timepoints. Survivorship checks were conducted in this way starting on November 9th and continued every other week until March 30th, when we performed our first spring inspection of colonies, and all colonies were opened, and survivorship confirmed.

Mass monitoring: To monitor the estimated consumption of honey stores within colonies over the winter, we tracked the mass of each colony starting on November 9th and continuing every other week until March 30th. We used an industrial crane scale (SAGA Perseus) with a handle attached to one end and a lever which hooks to the underside of the colony on the opposite end. We weighed the colony by lifting each side of the colony three times, taking an average weight of each side, and adding them together for total colony weight. Weighing the two sides of a colony with the lifting scale produces an imperfect biological mass, as all equipment is included and the position of the cluster can vary, however, within colony measurements are accurate and repeatable over time allowing for reliable measurements of change in mass. The scale features a “peak hold” setting which records and holds the maximum mass allowing the user to tilt the scale past center weight and obtain an accurate maximum weight for each side, this in combination with the average of three measurements per colony side reduces variation in weight due to user error. From this, we calculated the percent change in overwintering colony mass by subtracting the mass at each sample date from the starting mass of the colony on Nov. 9th, 2020. Specifically, percent mass was calculated as (1 - (current mass/initial mass) *100). The rate of mass decline was also determined for each colony using the slope of the linear trend equation for an individual colony change in mass over time (kg mass decline per sampling period).

The crane scale method to weigh hives is critical for this study, as other methods of weighing colonies usually require opening and disassembling hive equipment to weigh them independently, which would cause extreme stress during winter. We validated the crane scale weighing method by comparing the masses of 42 independent honey bee colonies, in the field during normal summer conditions. Each hive was measured with the crane scale method as described above and then by fully disassembling the hives and weighing the component pieces independently, as in Dolezal et al 2019. Using linear regression, we found that the crane scale method significantly predicted the weight determined by the more invasive method (F1, 40=136.7, p=<0.0001; S4 Fig), and the values were strongly correlated R2=0.77.

Frames sides of bees and capped brood area: Adult bee populations in colonies during spring buildup were estimated twice; on March 30th and April 12th. Populations were based on fractional estimates of sides of a frame covered in bees (i.e., “frame sides”) and capped brood area was estimated in each colony via photography following methods from Delaplane et al. 2013. In short, each frame was photographed, the area covered with brood was traced out using Photoshop, and the proportion pixels associated with capped brood was calculated. The proportion brood area was then converted to cm2 by multiplying the brood area by the area of the frame in cm2.

Collection of bee samples and lipid analysis: To measure colony lipid levels of nurse bees at the end of the experiment, we collected approximately 50 bees from a frame containing open brood during our spring inspection (March 30th, 2020). Bees were transported to the lab on ice and immediately stored at -80°C until further processing. Bees were processed via the protocol of Toth and Robinson as modified in Dolezal et al. (42). Approximately 15 nurse bees, by mass, were homogenized in liquid nitrogen, and approximately 0.3g of homogenate was subsampled and weighed. Lipid content was quantified via phosphor-vanilin spectrophotometric assay and lipid calculated as mg lipid/mg bee mass.

Landscape classification: We calculated the percent arboreal cover within 100m of colonies to better understand at what level variation in windbreak provided by tree cover contributed to the change in mass of colonies in covered vs control treatments. To calculate arboreal cover, we used Google Earth Pro to create a 100m radius buffer around the location of the colonies and then classified the number of pixels associated with the buffer in Photoshop. We then used Photoshop to create a layer that consisted of the arboreal cover that was within the buffer, calculated the pixels associated with this layer, and took a proportion of the total buffer.

Landscape scale land use surrounding each farm was quantified in ArcGIS, ArcMap 10.3.1 using a 1-km radius centered on the apiary location. Land use features were based on the US Department of Agriculture–National Agricultural Statistics Service cropland data layer for 2020 at a 30m × 30m resolution (https://nassgeodata.gmu.edu/CropScape/). Using the ‘histogram’ function in ArcMap, the proportion of all landscape feature classes were identified by counting pixels associated with each land category within the buffer. Land-use types were combined and categorized into four groups (cropland, developed, grassland, and woodland). 

Statistical Analysis: To compare colony temperatures across treatments, we created a mixed model analysis of variance in SAS using the ‘PROC GLIMMIX’ function. To meet the assumptions of normality, we used the natural log of the recorded temperature plus twenty-five as the response variable.  Treatment (i.e., control vs covered), sample date (i.e., calendar week for temperature), and their interaction were fixed effects in the model. Colony nested within site was used as a random factor. If significant main effects or an interaction were observed, then we conducted a post hoc analysis of least squared means with Tukey HSD adjustment for multiple comparisons to look for differences between treatments on individual calendar weeks. Colonies that died over the winter were censored from the analysis at the time point death was recorded.

We used the same model listed above to assess percent change in mass (i.e., change in mass from Nov. 9th, 2020), proportion mid-winter supplement consumed, frame sides of bees, capped brood area, and colony lipid percent. Any colonies that died, were censored from the analysis of percent change in mass at their time of death. Rate of mass decline was assessed using the same model as above with the exception that the fixed effect of sample date was replaced with apiary. Analysis of rate of mass decline only included colonies that survived the entire overwintering season.

To better understand the relationship between land use in the surrounding landscape and colony mass decline we conducted a model selection using multiple regressions with stepwise model selection in SAS with function ‘PROC REG’. Land use types cropland, woodland, grassland, and developed land were included in our model selections and required a p-value <0.15 for model inclusion. Because land cover types are inherently related to each other, we first ran a Pearson’s correlation using function ‘PROC COR’ in SAS to ensure that there were no collinearities among variables (Pearson’s correlation coefficient <0.8. To determine if there was a relationship with the percent arboreal cover within 100m of the apiary with the rate at which colonies lost mass we performed a linear regression in SAS using the ‘PROC REG’ function with rate of mass decline as the predictor variable and percent arboreal cover as the fixed effect. This analysis was performed for the covered and control treatments as well as at the apiary level.

To determine if colony overwintering survivability varied between treatments, we performed a Cox Hazard test in R using the ‘coxme’ package and function.

Tab - Overwintering Data

               Column

·        Yard - Each of the 8 yards used in the study

·        Colony ID – individual colonies used in the study

·        Treatment – the covered (called wrapped) or control (called unwrapped) treatments

·        Date – sample dates for each inspection

·        Survival – Whether a colony was dead or alive at an inspections. 0= alive, 1= dead

·        Raw mass (kg) – the mass of each colony in kg which was calculated by weighing each side 3 times with the crane scale, averaging the three weights per side, and taking a sum.

·        Percent change in mass over time – change in mass at each time point compared to Nov 9^th (considered time point zero).

·        Linear equation for growth – calculated by plotting the raw mass over time in excel and then displaying the linear equation.

·        Rate of growth (Mx) – by taking the Mx out of the linear equation from above (y=Mx+b)

·        Arboreal cover within 100m radius – calculated as described in methods.

·        Proportion mid-winter feed consumed with dead colonies filtered – we removed the proportion feed consumed for colonies that died over the winter. Overall proportion was calculated by taking the raw mass minus the mass of the cover board that was removed on Feb 2 when the feed was added. We then added the mass of the inserted feeder board which was 0.6kg. proportion was calculated in kg

·        Proportion emergency feed consumed, all colonies – same as above with the dead colonies included. Calculated in kg.

·        Frame sides of bees – calculated following methods from Delaplane et al 2013. Only calculated on 3/30/2021 and 4/12/2021

·        Queen sign – if we saw the queen or sign of the queen (eggs or young larvae) we assigned a y, if no sign was seen we assigned a n. Only measured on 3/3/2021 and 4/12/2021

·        Brood area (cm2) – the sum of the brood calculated as in the methods for each frame within a colony.

Tab – Preoverwintering Mites

               Column:

·        Yard: Colony: Treatment- as described above

·        October mites – mites levels in colonies from a 300 bee alcohol wash which was taken on October 5-8^th 2020.

·        November mites – mite levels in colonies from a 300 bee alcohol was which was taken on Nov 9^th 2020.

Tab – Colony Temperatures

               Column:

·        Yard: Colony: Treatment – same as above

·        Average temp – the average of each temperature taken every 256 minutes over the course of a calendar week.

·        Calendar week

·        sent-vs-live – sentinel are colonies that did not have live bees in them but were actually just empty colonies with a temperature monitor in them to calculate ambient temps. Live were target colonies within the experiment that were alive.

·        Log25- is the log transformation of the average temperature plus 25. LN(averagetemp+25)

Tab – tipping scale side experiment

               Column:

·        Colony- colonies used to in experiment that were located at Iowa State University during the summer 2019.

·        Tilted left – the weight of the left side of the colony using the tilt scale from this experiment in kg.

·        Tilted right – the weight of the right side of the colony using the scale from this experiment in kg.

·        Tilted total – the sum of the weight from the two sides in kg.

·        Classic scale weight – the weight in kg of colonies as measured on a postal scale following methods from Dolezal et al 2019.