Can colony size of honeybees (Apis mellifera) be used as predictor for colony losses due to Varroa destructor during winter?

For more than three decades, honeybee colonies (Apis mellifera) experience high losses during winter, and these losses are still continuing. It is crucial that beekeepers monitor their colonies closely and anticipate losses early enough to apply mitigating actions. We tested whether colony size can be used as early predictor for potential colony losses, in particular due to the parasitic mite Varroa destructor. V. destructor is one of the most important causes for these losses. Such early predictor for potential V. destructor induced losses is especially relevant as measuring V. destructor load in colonies is difficult and cumbersome. During three years, we monitored colonies with high and low V. destructor load from July until March of the next year. We found that differences in colony size were only visible after November, even though we lost almost all colonies every winter in the group with high V. destructor load. In the Northern hemisphere, November is considered to be too late for beekeepers to strengthen colonies in preparation for winter. We therefore argue that early-warning signs for potential colony losses due to V. destructor are urgently needed to allow beekeepers preventing winter losses. We discuss the role of precision apiculture to monitor the health and productivity of honeybee colonies.


Introduction
Several studies tested the presence and loads of honeybee pathogens or 74 colony health as predictive markers for winter honeybee colony losses (Dainat and cumbersome for beekeepers and scientists alike, especially when many 85 colonies should be monitored over time (Lee et al. 2010, Dietemann et al. 2013, Francis et al. 2013. Also, infection with deformed wing virus (DWV) and acute 87 bee paralysis virus (ABPV) in October are suggested as predictors for winter 88 losses (Johnson et al. 2009, Genersch et al. 2010, but that makes them rather 89 late-warning markers. Moreover, costs for beekeeping drastically increase when 90 these types of markers have to be applied at large scale. Therefore, we 91 question whether a simple and fast measure of colony size can be used to

105
(2012b) found that colonies that survived winter were larger in October than 106 colonies that collapsed. These two studies imply that colony size can be a 107 predictive marker of colony losses during winter, but the differences in colony 108 size between infested and uninfested colonies were detected relatively late in 109 the season.
In this paper, we address two questions: (1) whether the decrease in colony 111 size over time during the foraging season is related to V. destructor induced 112 losses during winter, and (2) whether the effects of V. destructor on colony size 113 can be found early enough in the season, namely before the winter preparation 114 period. To answer the first question, we compared colony size during summer 115 and survival during winter of colonies with high and low V. destructor load. We 116 did this by following colonies treated and untreated against V. destructor from 117 July until March of the next year. It appeared that most colonies with high V. 118 destructor load did not survive until March next year, so we could also use this 119 experiment to answer the second question.  For each year, the colonies in the group (N=10) with a low infestation of V.   compared to V-colonies. We noticed that many of the colonies not treated with 225 acaricides did not show any infestation in July or August (Table 3).
Colony size differed per year, month, and V. destructor infestation (Table 2, Fig.   228 2). However, differences in colony size due to V. destructor infestation did not in 229 general appear before December (Sidak post hoc test Varroa x Month). In one 230 year (2014) differences due to V. destructor levels appeared a little earlier in the 231 year, in November (Fig. 2c). Differences in colony size based on survival 232 (dead/alive in March) did also not appear during summer (Table 2, Fig. 2). In 233 October, the earliest difference in size was observed between the colonies that 234 did not survive until March and the ones that did (2014, Fig. 2f). In this paper, we investigate whether colony size can be used as early warning 248 predictor for colony losses during winter that are caused by the parasitic mite V. 249 destructor. In our experiment covering three years, we found that the colonies 250 with high load of V. destructor had a much higher probability of mortality before next spring than the colonies with low load. The differences in colony size 252 between the colonies with low and high load of V. destructor were only visible 253 after November, even in years when differences in mite levels already occurred 254 in July. Mere visual inspections of the colony sizes will therefore be insufficient 255 as predictive marker to prevent winter losses; late autumn is considered to be 256 too late for beekeepers to strengthen colonies in preparation of winter. We Additionally, levels above 6% showed more than 10% losses (Genersch et al. that beekeepers that are able to partly reduce mite infestation will be even much 307 less likely to find visual differences in colony size, but probably still have high 308 chances of winter losses.

309
There is clearly a need to search for colony traits as predictive markers, rather Vallone 2020) provide promising, non-invasive ways to measure colony health.

325
Experiments to test the potential of colony traits for early prediction of winter 326 losses to feed these approached are still needed, including the development of 327 devices or tools that allow easy and quick measurements of these colony traits.