HiveMind AI

Mites per 100 bees or mites per day: reading a varroa count you can trust

By HiveMind AI ·

Two beekeepers compare notes. One says her hive is "at 12": a sticky board, 36 mites over three days. The other says his is "at 4": an alcohol wash, 12 mites in 300 bees. His is the hive in trouble, and nothing in those two numbers tells you so.

Three of the common methods produce a rate, mites per 100 bees, against a sample you can hold in a measuring cup. The fourth produces a daily drop off a population nobody counted. No arithmetic turns one into the other.

Four ways to count, and what each one measures

Alcohol wash

Shake bees from one to three brood-nest combs into a tub, make sure the queen is not among them, scoop half a cup (about 300 bees) into a jar of alcohol, soap solution or winter washer fluid, and swirl for at least a minute. Strain through a screen, count, refresh the liquid and repeat; the Honey Bee Health Coalition guide says the second pass "increases the accuracy of the count".

The bees die, and the queen with them if you missed her. In return you get the most trustworthy denominator there is, because dead bees can be counted afterwards (UF/IFAS). The University of Minnesota Bee Lab calls the wash "more reliable than the powdered sugar roll test as fewer variables affect the efficacy".

Reliable is not automatic. In a 2025 Apidologie comparison of eight methods on 300-bee samples, one ethanol wash recovered 72% of the mites; three 30-second washes recovered 98%. The accuracy is in the repetition. Allow five minutes.

Sugar roll

Same 300 bees, about two tablespoons of powdered sugar, a vigorous minute of shaking, three to five minutes resting in the shade, then invert the jar and shake the sugar out over a white tray. Repeat, mist the sugar, count (HBHC; Penn State and USDA ARS publish shorter variants). The bees go back on the top bars.

"Harmless" is no longer a safe word: the Coalition's current guide says shaken bees "may or may not survive", and a 2025 mark-release-recapture study in the Journal of Apicultural Research found sugar-shaken workers died prematurely (summarised in Bee Culture). Cheaper than a wash, not free.

The roll is also technique-sensitive. Humidity or a heavy flow glues sugar and mites to the bees, and the Coalition calls the test invalid if the returned bees are not dusted dry. In the Apidologie comparison a single sugar shake recovered 78%, rising to 95% with two 15-second shakes and at least 10 g of sugar; a 2022 USDA study recovered 97 to 98% (Insects). Done well, it is a real per-100 count. Done in a hurry, it undercounts by an amount you cannot know, because the bees were never counted. Allow ten minutes.

CO2 injector

Same half cup, gassed until the bees stop moving, shaken 10 to 30 seconds so the mites drop through a screen, then tipped back once they wake. Fastest of the four, and the bees live. The independent evidence is thin: Randy Oliver's 2017 trial, checked by washing the same bees afterwards, recovered 40%, 70% and 58% of the mites at 10, 20 and 30 seconds, and his later methods review calls the method "inconsistent". We found no peer-reviewed head-to-head that includes CO2. Treat it as a per-100 reading that probably runs low, and never switch methods between checks on one hive.

Sticky board

An oiled or sticky board under a screened bottom, left 72 hours (Penn State, ARS). Count the mites, divide by the days: mites per day. Nothing dies and the hive stays shut.

What fell, though? Mostly mites that dropped as brood emerged. UF/IFAS puts it plainly: "mite fall is largely determined by the amount of emerging brood", and calls the method's accuracy "somewhat questionable". Ants carry mite bodies off (HBHC), and day-to-day drop on one hive varies widely (Scientific Beekeeping).

Why a board's number cannot be converted

A wash hands you both halves of a fraction: 12 mites, 300 bees, 4 per 100 (the Coalition's own worked example). A board hands you the top half only. The bottom half, the bees in the colony and the mites still sealed in brood, is unknown, and it changes with the season.

One authority tried the conversion. The UK National Bee Unit's fact sheet multiplies the daily drop by 30 in May to August, by 100 in March, April, September and October, and by 400 from November to February, marking the middle factor "approximate only". Ten mites a day is roughly 300 mites in July and roughly 4,000 in December. A number whose meaning moves thirteen-fold with the calendar is not a rate, let alone a rate per 100 bees.

The published "drop thresholds" say the same from the other side. USDA ARS treats at 30 or more per day; UF/IFAS at over 59 in 24 hours. Penn State's López-Uribe lab cites research pointing at 60 mites per board over three days, about 20 a day, then says outright: "there is not a reliable threshold for mites per sticky boards" (Penn State). A threefold spread among careful institutions is what a missing denominator looks like.

The Coalition files the board under less reliable methods, "useful to check mite population trends or as quick check to confirm treatment effectiveness". That is the right job for it: direction on one hive, over time, in its own unit. So HiveMind AI keeps the two units in separate lanes: a board reading is stored as mites per day, shown on the hive's timeline in those words, and never compared to the per-100 line that drives the alert. Not because we could not write the conversion. Because nobody can.

The thresholds, and why the experts disagree

Published action levels for wash, roll and CO2 counts, in mites per 100 bees:

Those last two are one institution, and the pairing is worth sitting with rather than explaining away. Penn State Extension's IPM page and the López-Uribe lab, at the same university, publish 2 and about 5. We have not mixed up a citation and neither of them has: that spread is what the disagreement actually looks like from the inside.

Nobody here is wrong. They keep bees in different climates with different stock and accept different odds. And each works from a sample that misses part of the population: adult-bee counts cannot see mites sealed in capped brood, which is why Lee and colleagues suggested doubling the adult figure to estimate the whole colony (J. Econ. Entomol. 2010). A threshold is a bet about what is behind the number.

HiveMind ships a default of 3 per 100, near the middle of that spread, labelled as a choice. Set your own extension service's figure in Settings → Alerts; we would rather you did.

Why the autumn line is lower

Every source above tightens as the season turns, and the reason is rarely explained. Two things happen at once.

The bees reared from late summer into autumn are the winter bees, the long-lived cohort (over 200 days) that carries the colony to spring (Locke et al. 2017). Reared beside a high mite load they pick up deformed wing virus as pupae, and no later treatment repairs that. Treat too late, Locke and colleagues write, and "the overwintering bees will have already been reared under Varroa-infested conditions and may be too ill-affected by virus infections to survive the winter, even if the mite treatment itself was effective".

Meanwhile the fraction moves against you. Brood rearing shrinks, so mites hidden in cells emerge onto a smaller adult population; the Coalition lists "decreased brood production, leading to a higher number of mites on adult bees" among the things that swing a reading. Jack and colleagues found colonies re-crossing 3 per 100 within three months of a summer or autumn treatment, against four to five months after a winter or spring one (Frontiers 2023).

So 2.5 per 100 in early September is not the same fact as 2.5 in May, and the Coalition's autumn line is 2. HiveMind's default has a "stricter from July to October" switch, set to 2 per 100, as a calendar stand-in for that post-harvest phase in temperate northern yards. South of the equator the equivalent window is roughly February to May; read the alert with that in mind.

The denominator is the whole measurement

Half a cup of lightly packed bees is about 300: UMN, Penn State and the Coalition use that equivalence, and UF/IFAS puts it at about 100 ml. Lee and colleagues built the 300-bee sample from one brood-nest frame into the plan every per-100 threshold assumes.

Now do the arithmetic on a half-filled scoop. Nine mites over 300 bees is 3 per 100. The same nine over 150 bees is 6 per 100. Enter 300 when you scooped 150 and you have halved the reading on a hive at twice the line: the wrong direction to be wrong in.

Take the bees from brood-nest combs, never the inner cover, the entrance or a honey frame (Penn State), let the older flyers lift off first, and count the actual bees now and then (HBHC). HiveMind's mite check defaults the sample to 300 and lets you change it, refuses a blank count rather than saving it as zero, and records one count per hive, because a wash measures one colony, not a yard.

What a count is good for

Three things.

A decision: this hive, this method, against the right seasonal line.

A trend: same hive, same method, roughly monthly (HBHC and Penn State say monthly; ARS says every three weeks in autumn). A board can carry a trend too, provided it is only ever compared with itself.

An efficacy check: a count before treatment and one after, same method, same unit. The Coalition expects a post-treatment wash under 2 per 100. Switch from a wash to a roll between the two and you have manufactured a result out of recovery rate rather than dead mites. What you treat with, and when, is regional: registered products and rules differ by state, province and country, so ask your extension apiculturist or apiary inspector.

Last, a count has a shelf life. With brood present, a mite population in non-resistant stock roughly doubles in about a month (Scientific Beekeeping). A reading older than about a month says a check is due, not what the load is today. HiveMind's alert treats a reading over 30 days old that way: it stops stating the figure in the present tense and asks for a re-check, and the HiveMind assistant is told to treat a stale count the same way, and never to compute efficacy for you.

Keep the two units apart, count the bees you sampled, and you have a number you can act on.

Recording one in HiveMind: Mite checks in the help centre covers the method picker, the two units, and the stale-reading alert.