Print Friendly

IPM 3 Fighting Varroa 3: Strategy – Understanding Varroa Population Dynamics

Page 6 of 7

Computer models

Speaking of further study, the way to find out if we truly understand mite population dynamics is to try to model them in a computer program. If our knowledge is correct, we will be able to predict what we actually observe in the “field.” If not, then back to the drawing board for fine tuning, or to the field to collect more data. Without a good model, our attempts at mite control are largely shooting in the dark! With a good model, we can aim accurately, and save our ammo. That is, by analyzing the models, one gets a much better grasp of which mite control strategies are likely to be effective.

The models take into account factors such as broodrearing season, proportion of drone brood, seasonal mite mortality, and reproductive rates of the mite, to name a few. I’ve cited the major modeling references at the end of this article, and the numbers I use in the following discussion come from these references. Note that the models may not apply well to your conditions! They are based upon many thousands of hours of tedious research and analysis by dedicated researchers, yet are still rather primitive. For instance, Stephen Martin’s excellent model only assumes drone brood rearing from May through August—which is very different from my California conditions! It also appears to underestimate the percent of drone brood that I see, and only has worker brood from April thru September. As the granddaddy of varroa modelers, Ingemar Fries of Sweden states: “No mathematical model will produce good results without reliable input data.”

You can play with modeling by downloading Gloria DeGrandi-Hoffman’s VarroaPop (Figure 3; see Resources). It’s not perfect yet, and I’ve spoken with Dr DeGrandi-Hoffman to clarify the parameters she used, but haven’t gotten the answers by deadline for this article.

Figure 4. An illustration of a Varroa Pop run. You can choose the parameters to plot. For this graph, I plotted only the mite population, not any of the bee population options. In this run, some of the variables I set were: Midwest weather, 1.1 mite offspring per worker cell, 3 per drone cell, 0.1% infestation in worker cells, 1% in drone, and moderate queen strength. Note that this model predicts that most of the mites would be in drone brood.

Although what catches our attention is mite damage to developing worker brood in fall, mite levels increase most rapidly when drone brood is being reared. Modeling shows us that mite reproductive increase, in a hygienic colony, is largely dependent upon four factors: (1) baseline reproduction in worker brood, (2) the availability of drone brood, (3) successful multiple cell invasions by each female, and (4) the restraint from daily mortality.


The average number of offspring produced per foundress (reproductive female) mite in worker brood is only a little over 1 viable female per reproductive cycle (about 17 days). Of these viable females, about 1/3 fall to the floor shortly after emergence (about 50% of them are dead, and the rest generally die within a day). Add to that other mite mortality factors, and the best that mites can do is to increase at a rate of about 2.4% per day on worker brood (doubling about once month). Point to note: if you have hygienic (VSH) bees that remove infested worker brood, this rate can be greatly reduced. Worker brood alone in a relatively hygienic colony is unlikely to generate dangerous mite levels. Serious mite increase requires the presence of drone brood to “prime the pump” to get any substantial mite production going in the worker brood (Martin & Medina 2004).


Mites have much greater reproductive success in drone brood—producing nearly 3 viable daughters per cycle. However, as you may have noticed if you’ve sampled drone brood with a cappings fork, there are often multiple mites in a single cell. Luckily for the bees, multiple infestation hurts the mites–when three foundress mites enter a drone cell at one time, their reproductive success is reduced by a third. So the varroa buildup in drone brood eventually slackens due to overcrowding of individual cells. Point to note: it is a core strategy of varroa management to minimize the number of drone cells in a colony, or better yet, to remove them on a regular basis during the spring, along with the mites within. This is especially true since for some reason hygienic colonies only remove infested worker brood, not infested drone brood.

This concept makes us rethink the effectiveness of attempting to kill the last few mites with a winter treatment. It’s counterintuitive, but our efforts toward going into spring with the lowest possible mite load may be misdirected—our energy may be better spent in managing drone brood. It would make sense that for every mite killed before spring there would be at least 12 fewer mites come fall, due to their exponential growth (Figure 5). However, this assumption does not take into account the restraining effect of multiple infestation in drone cells, as mentioned previously. By reducing the initial mite infestation, there is less mite-to-mite competition for pupal hosts, and the surviving mites are able to reproduce even faster and catch up! (Bieñkowska & Konopacka 2001). And this doesn’t even account for spring reinfestation from other colonies (Figure 6). So what I’m thinking, is that although the winter oxalic dribble is a good stopgap measure, in the long term we really need to look at ways to inhibit varroa reproductive success during the broodrearing period.

Figure 5. A simplified illustration of the buildup of mite populations over 180 days, starting with either 10, 100, or 200 mites. All curves are based upon a doubling of their population every four weeks. Note that all the curves are identical—just shifted time wise to the right at lower starting levels. Note that the mite population never reaches a danger level if initial mite level is very low. In reality, however, mites increase more quickly at low density, and slow down at greater mite density. Crown copyright 2005. From Managing Varroa.


In testing an early model’s predictions against reality in the field, Martin found that it underestimated the rate of reproduction in worker cells (Martin and Kemp 1997). I’m afraid that VarroaPop may also do so. If you play with VarroaPop, you’ll find that mite populations only increase if there is drone brood present, plus fairly high female mite “survivorship” (and therefore, multiple reproductive cycles by each female). This is a conclusion that Martin and Fries have come to—that the a female mite needs to average 2-3 reproductive cycles for varroa populations to grow at the pace that we see in the field. This is an aha! discovery. It means that a major weakness of the mite is the need for females to survive to reproduce between 2 and 3 times in order for the mite population to increase. Point to note: if we breed for bees with high grooming behavior, we’ll decrease the number of times that mites can survive to breed again. This factor largely accounts for the innate mite resistance of the Africanized bee.


On the subject of Africanized bees, they really do groom mites off their bodies well! They kick mites out of their colonies at nearly twice the rate that European bees do (Aumeier 2001). Point to note: aggressive grooming by the bees disrupts the mites’ behavior, and will likely increase the baseline daily mite mortality, especially if the hive has a screened bottom. Dislodged mites are normally able to quickly return from the floor by hitching a ride on a passing bee. If a mite falls through a screened bottom, it’s out of luck, and winds up becoming ant food. If you could increase daily mite mortality by just one percent by selecting for grooming behavior, you’d cut the mite buildup rate in half!


So far, I’ve spoken mostly of mites and mite population. But it isn’t usually the mites that kill a colony—it’s viruses. If you really want to understand the virus problem, read Sumpter & Martin’s (2004) paper on the dynamics of virus epidemics in honeybee colonies. You can skip the math—read the introduction and discussion. In short, there are at least 14 viruses endemic to honeybees. Before varroa, they were largely present at very low levels and “inapparent” (not causing symptoms). An occasional bee would show an “overt” infection, but die without necessarily spreading the disease. When varroa came on the scene, however, the mite suddenly became a new vector of the viruses, and at high enough mite levels, viruses become epidemic and cause colony collapse. Eventually, it’s likely that the bee, mite, and viruses will come to evolutionary terms, but that’s not the case yet.

The authors determined that since viruses are always present, the critical factor in predicting colony collapse was the mite population necessary to vector (spread) the virus to epidemic levels. Let me simplify the author’s mathematical predictions of those critical mite levels in Table 1. Note that the model’s predictions generally agree with the field-tested recommendations to keep mite populations below the 2000 to 3500 range.


Mite population models must account for invasion or “immigration.” Immigration is any inflow of mites from outside the colony—essentially by robbing of, and “drift” from, nearby collapsing feral or “managed” colonies. This factor is a “wild card,” and cannot be predicted by population dynamics, but it can reset the growth curve to a higher level when it occurs (Figure 6). Mite immigration is the tiger lurking in the bushes, and can overwhelm a colony in days! (I’m typing these words fresh from a beeyard of mine where this occurred—the lesson was painful).

Figure 6. Illustration of the effect of mite invasion (which mostly occurs through robbing of collapsing colonies). When a small number of mites are present at the start of the season and no mite invasion occurs, the mite population remains low. However, mite invasion early in the season causes the mite population to reach harmful levels much more quickly. Crown copyright 2005. From Managing Varroa.