A national survey of managed honey bee colony losses in the USA: Results from the Bee Informed Partnership for 2020–21 and 2021–22

Abstract In recent years, high losses of Western honey bee (Apis mellifera) colonies have been reported in the United States. Because honey bees are important to agricultural systems, it is critical to document when high losses occur and to explore patterns of loss among beekeeper subgroups. We used a voluntary, retrospective questionnaire to ask U.S. beekeepers (backyard, sideline, and commercial) about their colony loss during the 2020–21 and 2021–22 beekeeping seasons. We found that U.S. beekeepers lost 50.8% (38.0–63.1; 95% b.CI) of their colonies in 2020–21, the highest annual (year-long) loss reported to date. During the following year, 2021–22, beekeepers lost 39.0% (31.5–47.9; 95% b.CI) of their colonies, which is close to the average of previously published annual loss rates (40.9%). In contrast to previous years, backyard beekeepers (managing 50 or fewer colonies) had an elevated summer loss across both years and commercial beekeepers (managing >500 colonies) reported “weather” as an important cause of colony loss in the summer of 2021. Our results show that severe colony losses still occur periodically. From these data on colony loss rates, it does not appear that the health of honey bee colonies in the U.S. has improved since this survey began in 2008. We suggest that research should continue to focus on stressors that beekeepers most often perceive to be leading causes of loss—Varroa destructor mites, queen issues, and weather.


Introduction
To document and explain managed Western honey bee (Apis mellifera L.) colony losses in the United States, the non-profit Bee Informed Partnership (BIP) launched its Loss and Management Survey in 2011 as a continuation of the national Winter Loss Survey which began in 2007(vanEngelsdorp et al., 2007)).The survey initially focused on colony losses in the winter season (1 October to 1 April), when colonies depend on the extended survival of individual honey bees during a time of limited resource availability and opportunities to raise young bees (van Dooremalen et al., 2012).However, summer colony loss (1 April to 1 October) and annual colony loss (1 April to 1 April) have been reported since the 2012-13 survey year (Steinhauer et al., 2014).This and subsequent studies showed that substantial colony losses occur not only during winter but also during summer (Bruckner, Wilson, et al., 2023;Kulhanek et al., 2017;Lee et al., 2015;Seitz et al., 2015;Steinhauer et al., 2014).
To counterbalance losses, beekeepers can increase their number of colonies from within their operations through splitting-the act of dividing colonies into two or more smaller colonies (Connor, 2014)or by obtaining colonies from other beekeepers.However, high colony losses could potentially lead to declines in the total colony population if newly generated colonies are insufficient to compensate for losses (vanEngelsdorp & Meixner, 2010).Despite reports of high losses, estimates of the number of honey producing colonies in the U.S. have remained relatively stable since 2006, fluctuating between 2.3 and 2.8 million colonies (USDA-NASS, 2009, 2020).While it is important to have periodic information on the number of colonies in the U.S., it is also crucial to track colony loss rates because high colony losses can suggest vulnerabilities in the system that are not apparent when only considering annual reports of colony populations.
In past years of the survey, queen issues, Varroa destructor mites (Anderson and Trueman; hereafter "Varroa"), weather, starvation, and pesticides have been reported by beekeepers as important causes of colony losses (e.g., Bruckner, Wilson, et al., 2023).Researchers also view these as important causes of loss (Guzm� an-Novoa et al., 2010).We now understand colony loss as a multifactorial process in which concurrent and interacting stressors can lead to colony losses (Potts et al., 2010;Steinhauer et al., 2018).The prevalence and impact of specific stressors, such as Varroa and viruses, can vary with time of year (Faurot-Daniels et al., 2020;Martin, 2001), and may change over time (Martin et al., 2012).Therefore, it is valuable to document what beekeepers perceive as the leading causes of loss at different times of the year and to continue this monitoring over a period of years.
It has been suggested that commercially managed colonies encounter stressful environments and beekeeping practices which could increase losses (Simone-Finstrom et al., 2016).Some specific stressors of concern are exposure to pesticides and acquisition of pests and pathogens during pollination events (Alger et al., 2018;Frazier et al., 2015), as well as stress during long-distance colony trucking (Ahn et al., 2012;Simone-Finstrom et al., 2016).
A majority of colonies in the U.S. are transported across state lines, including to California for almond pollination (Goodrich et al., 2019).In past years of the survey, no association has been found between almond pollination or migratory beekeeping and elevated winter losses, and occasionally these activities have been associated with reduced winter losses (Lee et al., 2015;vanEngelsdorp et al., 2010).However, colonies that are transported across state lines and used to pollinate California almonds constitute a large and economically important segment of the U.S. honey bee population (Goodrich et al., 2019).Therefore, we continue to explore the potential association of inter-state movement and almond pollination on winter losses to monitor for problems among colonies that are managed in these ways.
The purpose of this study was to continue the long-term documentation of managed honey bee colony losses in the United States by reporting the 2020-21 and 2021-22 summer, winter, and annual loss rates.We reported colony losses nationally, by operation type (backyard, sideline, and commercial), and by state.We also documented what beekeepers perceive as the causes of summer and winter loss, and what they consider an acceptable rate of winter loss.Finally, we investigated possible associations of winter loss with transportation of colonies across state lines, and with the use of colonies to pollinate California almonds.

Survey
Following methods from previous years (Bruckner, Wilson, et al., 2023;Kulhanek et al., 2017;Lee et al., 2015;Seitz et al., 2015;Spleen et al., 2013;Steinhauer et al., 2014;vanEngelsdorp et al., 2007vanEngelsdorp et al., , 2008vanEngelsdorp et al., , 2010vanEngelsdorp et al., , 2011vanEngelsdorp et al., , 2012)), the 2020-21 and 2021-22 BIP Loss and Management Survey was a retrospective annual survey of beekeepers' colony losses and management practices.Participants were recruited directly via email (BIP mailing list, n ¼ 26,404), and indirectly by asking well-connected partners (e.g., beekeeping journals, beekeeping clubs, state beekeeping associations, and national beekeeping organizations) to advertise the survey to beekeepers.Additionally, snowball recruitment was used, whereby respondents were invited to promote the survey among their contacts.All responses were accepted through an online portal: The 2020-21 survey was hosted on a third-party service (www.SelectSurvey.com),whereas the 2021-22 survey was hosted on a domain owned by the Bee Informed Partnership (www.beeinformed.org).In both years, the online portals accepted responses from 1 to 30 April.The survey consisted of two distinct sections: "Loss" and "Management."Here we report on responses to the Loss section only.Other publications will report on beekeepers' responses to the Management section.
As in previous publications, we categorized beekeepers into three operational types based on the number of colonies managed on 1 October of each survey year."Backyard," "sideline," and "commercial" beekeepers were defined as those who managed 50 or fewer, 51-500, and more than 500 colonies, respectively.We refer to sideline and commercial beekeepers collectively as "large-scale" beekeepers.We also grouped respondents by whether they participated in specific beekeeping activities: moving colonies across state lines or using colonies for California almond pollination.Finally, we categorized respondents as single-state operations or multi-state operations depending on whether they managed colonies in one or several states, regardless of whether they reported moving colonies across state lines.
For each respondent, we determined the colony loss rate for three distinct seasons: summer, winter, and annual.Previously, the three seasons were defined as follows: summer (1 April to 1 October), winter (1 October to 1 April), and annual (1 April to 1 April) (e.g., Kulhanek et al., 2017).For backyard beekeepers, we retained these definitions because they appear to work well for the respondents.This also maintained consistency with past surveys.During focus sessions, large-scale beekeepers reported they often begin splitting colonies before the traditional 1 April start date for the survey's summer season.Therefore, the 1 April cutoff date is not suited to their seasonal calendar.According to these beekeepers, the moment preceding splitting typically represents the annual minimum number of colonies for large-scale beekeepers, so they consider the start of splitting to be the start of the beekeeping season.To reflect this, we defined the seasons for large-scale beekeepers based on this key beekeeping management action.If a large-scale beekeeper made their first split of the calendar year before 1 April (i.e., between 1 January and 31 March), their summer season started on the date of their first split.If no splits were made between 1 January and 31 March, then their start of summer defaulted to 1 April.Regardless of the timing of the start of summer, 1 October was always considered the beginning of the winter season for all beekeepers.
To determine colony losses for each beekeeping operation during the summer, winter, and annual seasons, we asked the same core questions as previous BIP Loss and Management Surveys (e.g., Kulhanek et al., 2017;Seitz et al., 2015).In brief, respondents reported the number of colonies they owned at the beginning and end of each season, the net number of new colonies made or obtained from within and outside their operation, and the number of colonies they sold or gave away.This allowed us to calculate the number of colonies lost.We also calculated the number of colonies "at risk" for each operation and season (i.e., the maximum number of colonies that could be lost during a season).As in previous studies, the number of colonies at risk was calculated by taking the starting number of colonies, adding any new colonies made or obtained, and subtracting any colonies sold or given away by an operation during a season (Steinhauer et al., 2014).We also asked beekeepers to report what they thought were the most "prominent cause(s) of colony death" for the winter and summer seasons.Beekeepers selected up to three options for both "causes of summer loss" and "causes of winter loss."Finally, we asked beekeepers to report what percentage of loss over the winter they would consider acceptable.
Following previously described methods, valid responses had to come from locations within the United States and have complete data to calculate the respondent's colony loss rate for either the summer or winter season.Additionally, we used an algorithm to identify and manually review potential duplicate responses (e.g., if someone stopped and later restarted the survey).Since some responses were valid for only one season, we made three datasets-for valid summer, winter, and annual responses.The annual dataset contained responses that were valid for both summer and winter seasons.Data validation, filtering, and per-respondent calculations of colonies lost and at risk were performed in Python 3.8.13(Python Software Foundation, 2022).

Statistics
We performed subsequent data preparation and statistical analyses in R 4.2.1 (R Core Team, 2021).We used tools in the tidyverse package for data preparation (Wickham et al., 2019), the stats and multcomp package in data analysis (Hothorn et al., 2008;R Core Team, 2021), the broom (Robinson et al., 2022) and gt (Richard Iannone et al., 2022) packages to prepare tables, and performed data visualization using the ggplot2 package (Wickham, 2016).
To aggregate data from individual respondents into loss estimates for the national honey bee population and loss estimates for subgroups of respondents (e.g., backyard beekeepers, beekeepers in a specific state, or almond pollinating beekeepers), we calculated the sum of the number of colonies lost for all beekeepers in that subgroup, and the sum of the number of colonies at risk for all beekeepers in that subgroup.To assign multi-state operations' colonies to individual states (for state loss estimates), we used the same method as the USDA National Agricultural Statistical Service Honey Report (USDA-NASS, 2022c) and previous survey papers (Steinhauer et al., 2014), wherein multi-state operations' colony numbers are counted in every state that they report keeping bees.However, we counted each colony only once for national estimates.We did not present results for states with fewer than 10 valid responses.
We calculated the loss rate for the national honey bee population and for each subgroup as the total number of colonies lost divided by the total number of colonies at risk.This corresponds to the "total loss" estimate (vanEngelsdorp et al., 2013) in previous surveys, in which each colony has the same weight in the estimated loss rate.We repeated this procedure to estimate the loss rate for each season we report on.We used a bootstrap method (n-outof-n, rep ¼ 1000) to obtain confidence intervals for these loss estimates (at the national level and for all smaller subgroups), which we abbreviate as "b.CI." The bootstrap is a method that allows us to estimate the uncertainty around statistics inferred from a dataset, without having to make assumptions about their distribution.The standard (non-parametric) bootstrap samples the original n observations from a dataset with replacement, creating b (number of bootstrap resamples, typically over 1000) pseudodatasets of size n, from which b statistics can be estimated.From the distribution of those estimates, we can calculate the bootstrapped version of the statistic (the mean of the bootstrapped statistics), as well as a nominal 1-a confidence interval for the original statistic (Chernick & LaBudde, 2011;Davison & Hinkley, 1997).We used the sample function from the base package (R Core Team, 2021) to resample our original dataset and create the bootstrapped datasets.
To compare colony loss between different subgroups of beekeepers (e.g., operation types, almond pollinators vs. not), we fitted generalized linear models with a quasibinomial distribution (quasibinomial glm) to account for overdispersion.As our general procedure to determine whether interaction effects or main effects of predictor variables were significant, we fitted a full model and a corresponding reduced model that omitted the term in question.We then performed an analysis of deviance (F test using the anova function) to determine whether the inclusion of the term was warranted, as one would when simplifying models by stepwise deletion (Crawley, 2013).For significance tests, we used an alpha level of 0.05.
To determine whether the national loss rate differed between summer and winter, we fitted a separate model for each year, with "season" as a main effect, and compared these models to intercept-only versions.We compared summer vs. winter losses but did not compare summer or winter losses vs. annual losses because an annual loss for a beekeeper is not independent of loss during summer or winter.To compare colony loss rates between operation types (backyard, sideline, commercial) within each season, we fitted a separate model for each combination of survey year and season (summer, winter, annual for each of 2020-21 and 2021-22) with "operation type" as a main effect.For pairwise comparisons between all three operation types within each year and season, we performed post-hoc tests using the glht function in the package multcomp (Hothorn et al., 2008) using the single-step method to adjust p-values.To investigate whether summer and winter losses differed within operation type, we fitted a new model for each combination of operation type and survey year (backyard, sideline, commercial for each of 2020-21 and 2021-22), including a "season" main effect.Again, we only compared summer against winter losses.To examine the association of colony migration or almond pollination with winter losses, we fitted a separate model for each year.We performed this analysis only on data for large-scale beekeepers (sideline and commercial) because few backyard beekeepers undertake these activities.Analyzing colony migration separately from almond pollination, we started with models that included main and interaction effects between "operation type" and "management practice" to allow for the possibility that the association between migration or pollination and colony losses could vary between sideline and commercial beekeepers.We did not find an interaction between operation type and migration on winter losses for either year (F ¼ 0.0597, p ¼ 0.808; and F ¼ 0.6455, p ¼ 0.423, respectively for the two years).The interaction between operation type and almond pollination on winter losses was not significant either year (F ¼ 0.2194, p ¼ 0.640; and F ¼ 1.0673, p ¼ 0.304, respectively).Therefore, we analyzed sideline and commercial beekeeper data together, and reported the association of each management practice and winter losses for large-scale beekeepers considered together.

Respondents
Backyard beekeeper respondents were much more numerous in our pool of respondents than sideline and commercial beekeepers.We estimated that the 2020-21 and 2021-22 surveys accounted for 6.7 and 10.9%, respectively, of the honey-producing colonies in the United States on October 1 of that year (USDA-NASS, 2021, 2022b).

Losses by operation type within season
During the 2020-21 survey year, backyard beekeepers' annual losses (56.3%; 54.5-58.1;95% b.CI) were significantly greater than those of commercial (50.2%; 33.1-63.6;95% b.CI) or sideline beekeepers (46.8%; 38.3-54.6;95% b.CI), while commercial and sideline losses were not significantly different from each other (Table 1; p-values in Table S2).During summer of 2020-21, commercial beekeepers had the highest estimated colony losses, followed by backyard and sideline (all significantly different).During winter, commercial beekeepers had the lowest losses, sideline losses were intermediate, and backyard beekeepers had the highest colony losses (p < 0.05 for all comparisons except sideline vs. commercial).
In the summer of 2021-22, commercial beekeepers had significantly lower losses than sideline or backyard beekeepers; sideline and backyard losses were not significantly different.During the winter of 2021-22, commercial beekeepers reported the lowest losses, sideline beekeepers reported intermediate losses and backyard beekeepers reported the highest losses (p < 0.05 for all comparisons).

State losses
Annual colony loss estimates for individual states for 2020-21 ranged from 20.7 to 85.8%, whereas annual state losses for 2021-22 ranged from 23.0 to 84.8% (Figure 2; Table S3).When multi-state operations were considered as a subgroup, these operations had an annual loss of 46.5% in 2020-21 and 38.8% in 2021-22.Annual, summer, and winter losses for individual states and multi-state operations can be found in Table S3.

Self-reported causes of summer and winter loss
When beekeepers listed the "most prominent cause(s) of colony death" in summer of both years, "queen issues" and "Varroa" ("Varroa mites and associated viruses") were their two most common responses, though the order varied by year and operation type (Figure 3).In summer of 2020-21, queen issues were the most common response in all operation types.Backyard and sideline beekeepers reported queen issues followed by Varroa in summer of 2021-22, while commercial beekeepers reported Varroa followed by queen issues.Even though it was the third-ranked response among commercial beekeepers, we noticed that "adverse weather" was reported by 41% of commercial beekeepers in summer of 2021-a large proportion compared to the previous 3 years (Bruckner, Wilson, et al., 2023).For winter 2020-21 and winter 2021-22, backyard, sideline, and commercial beekeepers all reported Varroa most commonly.For backyard beekeepers, weather was the second most common response in both years, while for sideline and commercial Orange and green colors represent loss rates above and below 30%, respectively.The grey color indicates that results are redacted for those states that had fewer than 10 respondents during a season.The survey year on the figure denotes the year the survey was conducted.For example, the 2022 survey year includes data for the summer of 2021 and the winter of 2021-22.MSO: multi-state operation; NCU: non-continental U.S. (Guam, Hawaii, Puerto Rico, and the U.S. Virgin Islands).Multi-state operations were counted in each state that they reported owning colonies but were also reported as a separate subgroup.
beekeepers queen issues were the second most common response in both survey years.

Losses by migratory practice and almond pollination
Large-scale beekeepers who moved their colonies across state lines had similar winter losses as those who did not-both in 2020-21 (F ¼ 0.037, p ¼ 0.848) and in 2021-22 (F ¼ 0.0072, p ¼ 0.933).Large-scale beekeepers who participated in California almond pollination had similar winter losses in 2020-21 as those who did not pollinate almonds (F ¼ 0.0874, p ¼ 0.768).In 2021-22, beekeepers who pollinated California almonds lost 23.1% of their colonies, while non-almond pollinators lost 39.3% of their colonies over winter.This estimated 16.2 percentage point difference was statistically significant (F ¼ 6.1455, p ¼ 0.015).

Discussion
Our findings for 2020-22 reaffirm the importance of monitoring colony loss rates in the United States and show that high losses still occur-though not every year.The national annual loss of 50.8% in 2020-21 was the highest estimated loss since 2011 when BIP started to monitor annual U.S. honey bee colony losses.In contrast, the estimated national annual loss rate in 2021-22 (39.0%) was comparable to previously published results (Bruckner, Wilson, et al., 2023;Kulhanek et al., 2017;Lee et al., 2015;Seitz et al., 2015;Spleen et al., 2013;Steinhauer et al., 2014;vanEngelsdorp et al., 2007vanEngelsdorp et al., , 2008vanEngelsdorp et al., , 2010vanEngelsdorp et al., , 2011vanEngelsdorp et al., , 2012)).Backyard beekeepers' summer losses appear to have risen since they were first surveyed: During the 2020-21 survey year, summer losses for the first time were nearly as high as winter losses.These elevated summer losses contributed to annual losses of above 55% for backyard beekeepers in The graph shows the proportion of beekeepers who identified each cause of loss as being in the top three causes of loss for their operation.Operation types are backyard beekeepers (1-50 colonies), sideline beekeepers (51-500 colonies), and commercial beekeepers (>500 colonies).Causes of loss are reported separately for the summer (yellow) and winter (blue) seasons.The points represent the numerical proportion of beekeepers reporting the cause of loss, and the error bars represent bootstrapped 95% confidence intervals (n-out-of-n, 1000 rep).ApiTreat: "Apicultural treatments"; Disaster: "Natural disaster"; EquipFailure: "Equipment failure (e.g., moisture, ventilation); Pollen: "Nutritional stress (pollen deprivation); QueenIssue: "Queen issues"; Scavengers: "Scavenger pests"; Weather: "Adverse weather"; ShedFailure: "Failure of environmental controls in sheds (wintering facilities)"; DK: "Don't know."2020-21 and 2021-22-the highest annual losses for backyard beekeepers since annual losses were first monitored.We noted a marked increase in the number of commercial beekeepers reporting "weather" as a cause of loss for summer of 2021 and a decrease in the number of backyard beekeepers reporting "starvation" as a cause of winter loss compared to previous years.Nevertheless, the beekeepers overall reported leading causes of loss that align with recent years: Varroa and queen issues (Bruckner, Wilson, et al., 2023).While it appears that colony losses in the United States are higher than they have been in past decades, a continued limitation to comparing losses in the U.S. to other parts of the world is that many other survey efforts only report winter losses, whereas annual losses would better allow worldwide comparisons.
Beekeepers of all operation types continue to experience rates of winter colony loss that are above what they consider acceptable, and are above government targets.Winter losses in both years were substantially higher than what beekeepers consider to be acceptable, which aligns with previous survey results (Kulhanek et al., 2017).Almost a decade ago, the Pollinator Health Task Force (2015) set a goal to reduce winter losses to no more than 15% by 2025.The winter losses of 34 and 24% estimated by this study are substantially above this target.Although we have not performed a full time-series analysis to analyze trends over time, we do not see evidence that honey bee colony loss rates have decreased since annual losses were first recorded in 2011, which indicates continued poor health of honey bee colonies in the U.S. High rates of loss by individual beekeepers can be costly and discouraging for those who keep bees for pleasure and can have grave economic consequences for professional beekeepers.
Not only was the national annual loss in 2020-21 the highest reported to date, but the estimated colony loss rate during each season during 2020-21 (summer 31.1%,winter 33.5%, annual 50.8%) was higher than the mean of previously reported losses for those seasons (summer 23.5%, winter 29.0%, annual 40.9%) (Bruckner, Wilson, et al., 2023;Kulhanek et al., 2017;Lee et al., 2015;Seitz et al., 2015;Spleen et al., 2013;Steinhauer et al., 2014;vanEngelsdorp et al., 2007vanEngelsdorp et al., , 2008vanEngelsdorp et al., , 2010vanEngelsdorp et al., , 2011vanEngelsdorp et al., , 2012).It appears that in 2020-21, a summer of high losses was followed by a winter of high losses, resulting in a high loss year.While the annual losses in 2020-21 were high, estimates from USDA-NASS indicate that the number of honey bee colonies in the U.S. decreased by 5% from April 2020 to April 2021 and then returned to previous levels (USDA -NASS, 2021-NASS, , 2022b)).From these data, there is no indication that the national honey bee population is in decline.It is important to note that our estimates of national colony loss rates and any estimate of the number of colonies in the U.S. mainly reflect the state of the commercial beekeeping sector because commercial beekeepers manage most colonies in the U.S.
Although commercial beekeepers keep most of the national population of colonies, most beekeepers in the U.S. are backyard and sideline beekeepers who report different seasonal patterns of colony loss than their commercial counterparts (e.g., Bruckner, Wilson, et al., 2023).Backyard beekeepers tend to have higher annual losses than commercial beekeepers (e.g., Bruckner, Wilson, et al., 2023), with sideline beekeepers often but not always having intermediate levels of loss.Furthermore, backyard beekeepers tend to lose substantially more colonies in the winter than they do in the summer, but this pattern changed in the years reported here.While backyard beekeepers' summer losses in previous years were usually below 20% (Bruckner, Wilson, et al., 2023;Kulhanek et al., 2017;Lee et al., 2015;Seitz et al., 2015;Spleen et al., 2013;Steinhauer et al., 2014), in 2021 they had their highest summer loss so far (35%), and for the first time their summer losses were nearly as high as their winter losses.This apparent increase in summer mortality among backyard beekeepers deserves further investigation to see if it continues-and if so to determine what may be causing an increase in summer loss.
Backyard beekeepers as a group attributed summer and winter colony losses to a broader array of factors than did sideline and commercial beekeepers-and their leading responses included Varroa, queen issues, starvation, weather, and "don't know."Commercial beekeepers agreed more with one another in that they most commonly cited two causes of loss: Varroa and queen issues.The broader set of responses from backyard beekeepers may suggest that backyard colonies are subject to a broader set of stressors.Through actions that systematically manage stressors (e.g., sugar feeding) commercial beekeepers' colonies may experience a reduced set of stressors.Supporting this idea, we know that commercial beekeepers tend to manage their colonies in ways that align more closely with expert recommendations (Steinhauer et al., 2021).Additionally, commercial beekeepers often winter their colonies in locations with a mild climate (or indoors) and likely develop their operations in locations that are favorable for colony survival.In contrast, the home location of a backyard beekeeper will strongly influence where they choose to keep their bees, presumably meaning that some of these locations are less suitable than those experienced by commercial colonies.If commercial beekeepers do in fact keep their colonies in more favorable locations, this along with more intensive management could lead to fewer major causes of loss and lead to lower loss overall among commercial beekeepers.Other survey efforts have also found that larger beekeeping operations tend to suffer lower losses (Brodschneider et al., 2016;Gray et al., 2023) and have also pointed to greater experience among larger operations.Others have also noted the fundamental connection between colony losses and operation size: operations with lower losses would be more able to grow larger.Although commercial beekeepers' leading self-reported causes of summer loss were queen issues and Varroa, we noted that in summer of 2021, more of them attributed summer losses to "weather" than in the past.In the summer of 2021, 41% of commercial beekeepers cited weather as a cause of summer loss, compared to only about 10% in the preceding three years (Bruckner, Wilson, et al., 2023).This increased attribution of summer losses to weather coincides with the widespread drought conditions in the Northern Great Plains and Western U.S. (NOAA-NCEI, 2022)-important summertime beekeeping areas for migratory beekeepers (Smart et al., 2016).Recent studies further corroborate the potential importance of weather to both summer (Calovi et al., 2021) and winter (Insolia et al., 2022;Overturf et al., 2022) managed honey bee colony losses in the United States.
Our results show a broad consensus across all operation types of U.S. beekeepers that Varroa and associated viruses are the leading cause of colony mortality during winter, and this aligns with our understanding that Varroa infestation has various negative effects on individual bees (e.g., Bruckner, Straub, et al., 2023) and is an important driver of colony mortality (e.g., Chauzat et al., 2010;Genersch et al., 2010;Guzm� an-Novoa et al., 2010;Steinhauer et al., 2018).However, beekeepers' perception of Varroa as a cause of loss varies with operation type.About 60% of commercial and sideline beekeepers cited Varroa as a prominent cause of winter loss, but only 37-43% of backyard beekeepers did.This difference may indicate that backyard beekeepers are underestimating the effects of Varroa on their colonies.Because large-scale beekeepers more systematically manage non-Varroa stressors, it may be clearer to these respondents that many losses are caused by Varroa and associated viruses, which are difficult to control even with the tools that currently exist (Jack & Ellis, 2021).The emergence of amitraz resistance in Varroa (Rinkevich, 2020) may be further limiting the usefulness of the existing chemical varroacides, and is one possible contributor to the recent high losses.The report that Canadian beekeepers had an unusually high winter loss of 45.5% in the winter of 2021-22 (CAPA, 2022) ) indicates that the phenomenon of periodic elevated losses is not isolated to the United States and could implicate overlapping causes such as Varroa.
Like in previous BIP surveys, we did not detect an association between colony movement or almond pollination and increased winter losses among largescale beekeepers.In fact, for the 2021-22 season, almond pollinators lost fewer colonies than non-pollinators.However, we note that we analyzed the association of these activities and losses only to rule out major negative effects of either management practice.A limitation of the single-variable modeling approach is that the actual effects of almond pollination may be confounded by specific characteristics of migratory operations or almond pollinators that affect colony losses but were not included in the analysis.Still, the results presented here and in previous surveys argue against catastrophic direct impacts of either interstate movement or almond pollination.This suggests that the downsides of migration and almond pollination (e.g., shipping stress, pesticide exposure) (Ahn et al., 2012;Simone-Finstrom et al., 2016;Wade et al., 2019) are balanced by the benefits of migration and almond pollination (e.g., favorable climate, availability of plentiful forage), and/or by the benefits of other management that migratory and almond-pollinator beekeepers perform.One potential hypothesis for the large apparent benefit of almond pollination in 2021-22 is the warm weather favorable for pollination (and bee foraging) in the early part of the California almond bloom in 2022 (USDA-NASS, 2022a).If less favorable weather occurred in the locations where non-almond pollinator large-scale operations were based, this could have increased losses among those non-almond pollinators.
It is not straightforward to place U.S. colony losses in the context of present-day losses in other regions (Bruckner, Wilson, et al., 2023).To enable comparisons across countries with differing climates, we think it is essential for studies to report annual (year-long) losses, yet many survey efforts report only winter losses (e.g., Brodschneider et al., 2010;Gray et al., 2023;Zee et al., 2014).The focus only on winter losses probably reflects the dominance of winter losses in cooler areas of Europe.However, our results from the U.S. in this and previous BIP surveys show that summer losses can be substantial; this is corroborated by results from Uruguay that showed similar summer and winter losses (Ant� unez et al., 2017).Additionally, results from a survey in South Africa showed high losses during a 6month period characterized as the "active season" (Pirk et al., 2014), and results from a survey in Saudi Arabia found losses occurring during all seasons of the year (Albarrak & Gray, 2023).Even in Europe, we see some evidence that summer losses are currently underappreciated.While few studies report both summer and winter losses, a field study reported both percent summer loss and percent winter loss across 17 European countries ranging from Scandinavia to the Mediterranean (EPILOBEE Consortium et al., 2016).From these results, we calculated the ratio of winter losses to summer losses for each country and found a clear effect of latitude on the relative impact of summer and winter losses (Table S4).In the four countries with the highest relative impact of winter losses (Lithuania, Latvia, Poland, and Slovakia), winter losses were 12.9-72.3times as high as summer losses, while in the four countries with the lowest relative impact of winter losses (France, Italy, Spain, and Greece), winter losses were estimated to be only 1.3-2.3times as high as summer losses.These calculations may still underestimate the relative impact of summer losses in the Mediterranean region given that: (1) the study tracked already established colonies beginning in autumn (i.e., not nucs/ splits), and (2) the colonies were monitored for an abbreviated summer period (i.e., not all the way to the subsequent autumn).Given these facts, we think monitoring colony losses only during the winter season may limit the comparison of colony loss rates even between countries within Europe.In the tropics and subtropics, or in areas with dry summers, phenomena such as precipitation may be more important in determining the active season for honey bee colonies (Requier et al., 2018), and in some cases, there is not one season in which losses are concentrated (Ant� unez et al., 2017).
To bridge the differences in seasonality and in seasons during which the majority of losses occur, we think the best option is to record losses that occur over an entire year-accounting for increases and decreases from causes other than colony loss (Steinhauer et al., 2014).Asking beekeepers about their annual losses does not preclude gaining more seasonal resolution by also asking them about specific seasons of interest (e.g., summer and winter seasons in the BIP surveys), and it does not preclude defining the start or end dates in connection to biological events or management actions (e.g., end of winter preparations in the COLOSS surveys; the date that large-scale beekeepers made the first split of a calendar year in this BIP survey).When there may be large differences in the length of a season of interest, such as a beekeeper-defined winter season (van der Zee et al., 2012), it would aid interpretation of results if data on the start and end date of the season were collected and descriptive statistics on its length could be presented across different subgroups of respondents.

Conclusions
The 2020-21 and 2021-22 Bee Informed Partnership Loss and Management Survey revealed that high losses of honey bee colonies still occur in the United States.Despite public attention to the importance of pollinators and investment in honey bee research, loss rates do not indicate an improvement in the health of honey bees over the past 10-15 years.Observational studies like this survey cannot on their own conclusively demonstrate causal relationships, but this research points to multiple factors that beekeepers perceive as important causes of colony loss-Varroa, queen issues, and weather.We hope that research efforts to better understand these problems can lead to advances in honey bee breeding, product development, and management practices that can be broadly applied to improve colony health.We should continue to monitor colony losses to gauge progress on colony health.

Figure 1 .
Figure 1.Colony loss rates (%) of Apis mellifera honey bee colonies in the United States.Panes specify losses for (A) all operation types; (B) backyard beekeepers (1-50 colonies); (C) sideline beekeepers (51-500 colonies); and (D) commercial beekeepers (>500 colonies).The points represent the percentage of colonies lost, and the shaded bands and error bars represent the bootstrapped 95% confidence intervals (n-out-of-n, 1000 rep).Yellow, blue, and red represent summer, winter, and annual seasons, respectively.The survey year on the figure denotes the year the survey was conducted.For example, the 2022 survey year includes data for the summer of 2021 and the winter of 2021-22.Data for 2008-2020 are from previous surveys.

Figure 2 .
Figure 2. Colony loss rates (%) of Apis mellifera honey bee colonies in U.S. states and territories in 2020-21 and 2021-22.Orange and green colors represent loss rates above and below 30%, respectively.The grey color indicates that results are redacted for those states that had fewer than 10 respondents during a season.The survey year on the figure denotes the year the survey was conducted.For example, the 2022 survey year includes data for the summer of 2021 and the winter of 2021-22.MSO: multi-state operation; NCU: non-continental U.S. (Guam, Hawaii, Puerto Rico, and the U.S. Virgin Islands).Multi-state operations were counted in each state that they reported owning colonies but were also reported as a separate subgroup.

Figure 3 .
Figure 3. Causes of Apis mellifera honey bee colony loss as reported by United States beekeepers in 2020-21 and 2021-22.The graph shows the proportion of beekeepers who identified each cause of loss as being in the top three causes of loss for their operation.Operation types are backyard beekeepers (1-50 colonies), sideline beekeepers (51-500 colonies), and commercial beekeepers (>500 colonies).Causes of loss are reported separately for the summer (yellow) and winter (blue) seasons.The points represent the numerical proportion of beekeepers reporting the cause of loss, and the error bars represent bootstrapped 95% confidence intervals (n-out-of-n, 1000 rep).ApiTreat: "Apicultural treatments"; Disaster: "Natural disaster"; EquipFailure: "Equipment failure (e.g., moisture, ventilation); Pollen: "Nutritional stress (pollen deprivation); QueenIssue: "Queen issues"; Scavengers: "Scavenger pests"; Weather: "Adverse weather"; ShedFailure: "Failure of environmental controls in sheds (wintering facilities)"; DK: "Don't know."
bLoss estimates are presented for (A) all U.S. beekeepers; and (B) by operation type.Operation types are backyard beekeepers (1-50 colonies), sideline beekeepers (51-500 colonies), and commercial beekeepers (>500 colonies).n: number of valid respondents.Square brackets denote bootstrapped 95% confidence intervals.In (A), different letters show statistical significance between summer and winter within each year.In (B), different letters show statistical significance between operation types within each season for each year.For p-values of these comparisons, see Table