Springtime Treatment Thresholds
One of the mistakes I most enjoy repeating is the failure to adequately reduce mite loads coming out of winter. It makes for a flat year of honey production and a dead colony before the second winter ends. With a focus on spring management, two questions come to mind: What is a and what is the treatment threshold.
There is a clinical definition, but in the context of varroa management what happens at a 2% infestation that doesn’t happen at 1%? Left untreated, both can be expected to produce a harvest before fall collapse, so the threshold does not appear to be a matter of imminent danger. This is supported by Randy Oliver's observation (scientificbeekeeping.com) that mite-vectored viruses don't have a noticeable, colony-level impact until above 5% infestation (refer to notes on the mite model). Still, recommended limits have fallen considerably lower.
This downward trend has always been, giving credence to the theory that varroa treatments breed stronger mites and weaker bees. In many places, one mite appears enough to issue a colony its death sentence. The infestation level simply reflects time remaining on death row. The beekeeper can stay the execution by knocking the mites back, but there has been no reason to believe a common queen will produce a resistant colony. At the same time, treating ten thousand bees infested with a single mite seems unreasonable. Enter the treatment threshold, which is ridiculous in its own right. The beekeeper is advised not to treat, not to treat... right up until the point treatment is overdue.
The questions remain: What is the springtime treatment threshold, and what does it mean? Using a combination of personal experience and mite model simulations, I offer the following answers:
1. A springtime treatment threshold is the maximum allowable mite drop in April that will prevent exceeding a 3% infestation by August (at which point supers should be off and the critical summertime treatment applied). This is because once infestation crosses 3% in late summer, it will continue to climb so fast, so high that the beekeeper is unlikely to regain control. PULL UP!
2. The springtime treatment threshold is zero to one mite in a wash of three hundred bees, because modeling predicts that if more than one mite drops in April, infestation will exceed the 3% threshold by August, and in some cases 3% will be exceeded if just one mite drops in April (0-1 mites thus allowing thresholds to accommodate small differences in risk tolerance and/or sampling proficiency).
Let us compare this to thresholds published by the Honey Bee Health Coalition. Recommendations apply to either an alcohol wash or sugar shake of three hundred bees from a brood frame. The “acceptable” mite load during springtime is below 1% (0 to 2 mites). After that, the advice given depends where on the page one looks:
A bit murky is the advice given above. This may be the nature of mite management, but these recommendations leave the beekeeper damned if she does and damned if he don't. When it comes to Varroa mites, beekeepers are much sooner damned if they don’t. Simplify these parameters with one rule:
When there is a reasonable expectation that treatment will be needed within a week or month, consider any allowance for delay as an immediate need to treat.
The HBHC affirms this recommendation to avoid delay elsewhere in its publication. The result is an unconditional threshold below 1%. Any reward for hesitation is unlikely to justify the risk. The question now becomes: What is the probability that a sample of three hundred bees from a brood frame will drop at least three mites and alert the beekeeper of the need to treat?
Using a binomial probability calculator (Oliver 2020) the answer is 58%, provided the sampling is performed perfectly. This is fully unacceptable. And when is sampling ever perfect? This is low enough to suggest that probability has already been factored. However, the very meticulous mite model developed by Oliver demonstrates in head-aching detail the autumn collapse of an untreated colony having a one percent spring infestation.
Now consider that a cup of three hundred bees looks remarkably similar to a cup of 265 bees. Don’t expect to notice a difference. When the sampling cup is thirty-five bees short, the probability of detecting a 1% infestation falls below 50%. Likewise, when the cup runneth over, infestation will be overestimated.
What happens when a frame without brood is sampled? This is a distinction that isn't made in the mite model, which suggests it doesn't matter what frame from the brood box is chosen. Support comes from research that has indicated a random distribution of mites in the hive (Oliver 2020). However, other research has found mites aggregated on brood combs (Lee 2010). So, which is it?
The difference appears to be time of year, as pheromones of nurse bees and foragers begin to overlap later in the season (Watkins de Jong 2020). It is uncontroversial that mites prefer nurse bees, and as brood area declines ahead of winter, an increasing number of nurses are displaced from the brood nest (Mattila 2001). A similar phenomenon occurs in highly infested colonies. The mite severely limits a bee's capacity to produce the larval food, resulting in an early departure of infected bees from the brood area (Zanni 2018). These scenarios serve to disperse mites throughout the hive and are common from late summer onwards.
Conversely, an expanding colony is characterized by a well-defined division of labor and the absence of surplus nurse bees (Johnson 2009). This helps keep mites concentrated on brood frames during spring buildup. Unlike similar studies, (Lee 2010) analyzed mite distribution earlier in the season and found a 33% higher infestation of bees on combs having brood, which she considered an underestimation. Even at this conservative figure, the probability for a perfect alcohol wash to detect a 1% springtime infestation in a sample of three hundred bees from a non-brood frame is 40%.
Expect a sugar shake to drag these numbers further down. Even if the alcohol wash is perfect, who would bet their colony on a 58% chance? These numbers are not acceptable, so the simple solution is to round up and call two mites the same as three. Probability of detecting a 1% infestation now jumps to 80%, provided the sample is taken from three hundred bees on a brood frame. If sampling a frame without brood, probability increases to only 66%, which my grade school teacher still considered a failure. It must be concluded that for the purpose of making springtime treatment decisions, sampling a frame without brood is not supported. This is in agreement with HBHC guidance, as well as conclusions drawn by Marla Spivak (2010).
Is 80% good enough? If the alcohol wash is not perfect, or if performing a sugar shake, the probability is 70-something %. At risk is your colony. What do you stand to gain? The reward is a one-time avoidance of unnecessarily contributing to the development of stronger mites and weaker bees, in addition to preventing any direct harm caused by the poison itself. These are worthwhile goals. They are meaningful benefits that complicate the treatment decision.
Interestingly, despite different methodologies, we arrive at the same springtime treatment threshold of zero to one mite, depending on risk tolerance and sampling proficiency (or sampling choice - sugar shakers should lean toward the stricter threshold).
Some important points to keep in mind:
Sampling is required to learn the degree of infestation (provided the colony is strong enough).
Choose a frame having eggs or brood in any stage of development to ensure threshold reliability.
Measure 1/2 cup of bees as accurately as practical (count the bees after an alcohol wash).
Use a treatment threshold (allowance) of zero to one mite (depending on factors as given).
If sampling monthly, treat whenever infestation fails to decline, even if below threshold (see below).
CAUTION: If sampling every month in spring, regardless of threshold, the same infestation in consecutive months indicates the need to treat, even when results are below threshold. Although it may seem that mite levels are well managed, consider that modeling predicts autumn collapse of the following colony:
When the colony grows at about the same rate as the mite population (as happens in spring), infestation remains about the same (Randy's Varroa Model). This creates the illusion that mites are under control when the colony has done little or nothing to slow their growth. No further information is needed to make a treatment decision.
SPRINGTIME TREATMENT THRESHOLDS
IMPORTANT: What about treating in spring without sampling first? The recommendation against this practice is increasingly difficult to defend, as thresholds continue to fall. There even exists a range of circumstances in which sampling may lower colony survival by inducing a false sense of security in a beekeeper who is otherwise ready and willing to treat (personal observation).
It is unlikely for a colony to emerge from winter with a mite load below 1%. This is because in many cases, there may not be a broodless period that allows for a vapor treatment to kill all the mites (Avitable 1978). Even if there is, the window is likely to be short and tricky to estimate. Then the temperature has to allow for a loose cluster. More importantly, the mite load has been shown to increase in winter, even if no brood is reared at all, as mites detect and easily transfer from dying bees onto healthy ones (Bowen-Walker 1997).
In other cases, a colony may be too small to sample. That may also mean it is too small to treat, but waiting for it to grow could be their final demise. In this situation, it is probably best to just treat. Another such scenario is anytime before a new package begins capping brood or upon taking possession of a nuc that has not been verifiably treated.
Special circumstances aside, sampling before making a treatment decision is the best option. Proficiency in estimating the mite load is an important beekeeping skill, not only for avoiding unnecessary treatments but to monitor infestation trends under various circumstances, determine treatment efficacy, verify high infestation, perform a colony autopsy and identify resistance.
Finally (and nearly needless to say), although based on scientific discovery, the recommendations given here are one man's opinion.
Randy's Varroa Model; Randy's Varroa Model - Scientific Beekeeping
Honey Bee Health Coalition; Honey Bee Health Coalition Tools for Varroa Management
Binomial Probability Calculator; Binomial Probability Calculator (stattrek.com)
Re-Evaluating Varroa Monitoring: Part 2; Oliver (2020) Questions on Sampling Hives for Varroa - Scientific Beekeeping
The Role of Varroa Mite Host Selection on Forager-Mediated Mite Migration; Watkins de Jong (2020)
Timing of production of winter bees in honey bee colonies; Mattila (2001)
Reduced nursing by mite-infested bees depends on accelerated behavioral maturation; Zanni (2018)
Division of labor in honeybees: form, function, and proximate mechanisms; B Johnson (2009)
Practical Sampling Plans for Varroa destructor; K Lee (2010) BA213302.pdf (extension.org)
Standardized Sampling Plan to Detect Varroa Density in Colonies and Apiaries; M Spivak (2010) (PDF) Standardized Sampling Plan to Detect Varroa Density in Colonies and Apiaries (researchgate.net)
Brood Rearing in Honeybee Colonies from Late Autumn to Early Spring; A Avitable (1978)
Distribution of Varroa on overwintering honeybees and changes in levels of parasitism; Bowen-Walker (1997)