It’s getting cold at night. Should I harvest before the frost?
Usually not yet — and the thing that actually ends an outdoor grow in October is normally not the thermometer.
By Curtis Olson — retired software developer, home grower, and the builder of the device this site is about. More about me and why I built it. · Last reviewed
The short answer
A run of cold nights — nights in the 40s °F, around 5–9 °C — is not a harvest deadline. Cold slows a plant down; it takes a freeze to end it, and those are two different weather events that arrive under different conditions.
What you should actually be watching in the last few weeks outdoors is water — rain, dew, and how many hours your buds stay wet — because that is what takes outdoor crops in autumn far more often than temperature does. The cold matters mostly because of what it does to the humidity, not because of what it does to the plant.
So the honest sequence is: work out whether you are facing a cold night or a real freeze, then judge ripeness on its own terms, and let the wetness — not the forecast low — be the thing that makes you chop early.
A cold night and a frost are not the same event
This is the distinction the question usually hides, and getting it straight settles most of the worry.
A cold night is just a low temperature. Nothing freezes. The plant slows down, sits well below the temperature where it photosynthesises best, and carries on.
A radiation frost is what people normally mean by “frost.” It happens on clear, calm nights, when the ground and everything on it radiates more heat away to an open sky than it gets back. The standard reference on the subject — the FAO’s frost-protection manual — describes these as occurring with a clear sky, little or no wind, a temperature inversion and a low dew point. The giveaway is the combination: cloudless, still, and dry.
An advection freeze is the other kind: a mass of cold air moving in, usually with cloud and a decent wind. Different animal, and the things that protect against a radiation frost mostly do not help.
Why this matters to you tonight: a clear, calm, dry night at 3 °C is a more serious proposition than a breezy, overcast night at 1 °C, even though the second number looks worse. And wind, which feels like the dangerous ingredient, is the one that reduces radiation-frost risk, because it mixes the air instead of letting the coldest layer pool at the surface.
What happened when researchers actually flowered cannabis in the cold
There is one study that speaks to this more directly than anything else I have found, and it is worth knowing exactly what it did.
A 2025 paper in the Journal of Cannabis Research took a day-neutral CBD cultivar, grew it in a greenhouse at around 24 °C until a couple of weeks after flowering started, and then moved the plants into growth chambers held at constant temperatures ranging from 0.5 to 22 °C — that is the temperature around the clock, not just an overnight dip — and carried them all the way to harvest.
The plants at the bottom of that range finished. They produced inflorescence that could be dried and weighed, and CBD that could be measured. A whole flowering period spent at a constant 0.5 °C, which is colder than the 4 or 5 °C that a 40 °F night amounts to, did not kill them.
Two honest caveats, because this is a single study and I would rather you had its limits than a slogan. It is one cultivar, day-neutral and bred for purple, in growth chambers. And 0.5 °C is above freezing — nothing in that experiment addresses ice forming inside plant tissue. It tells you a great deal about cold and nothing at all about a hard freeze.
A separate Cornell study points the same way from a different angle. It put two hemp cultivars through repeated cold treatments and varied how often and how hard they came, and the authors’ summary of the result is that few effects of those sequential cold treatments were noted. Repeated cold events, in other words, are not cumulative damage in the way a run of cold nights on a forecast makes you feel they must be.
The usual answer on the forums is right, for a reason that turns out not to hold
Ask this question anywhere and the reassuring reply comes back in one shape: it’ll be fine, it’s been outside all season, it’s hardened off.
The conclusion is almost certainly correct. The reason attached to it is the part I cannot support, and I think that is worth knowing rather than glossing.
The Cornell study above tested exactly this. It included a cold-acclimation pretreatment — the experimental version of “it has been out there all season” — and the finding was that the effects of cold were not moderated by acclimation, and that the acclimation conditions themselves tended to have negative effects across many of the responses measured, including lower total CBD and total THC than in the plants that had not been acclimated.
That is not nothing. Hardening in this plant is not the shield the folklore says it is, and there is at least a hint that the process of getting hardened costs something.
To be fair to the other side: an earlier study did find that cold acclimation produces real molecular changes in Cannabis sativa — measurable shifts in the chromatin state and expression of cold-responsive genes. So the machinery is there and it does respond. What the whole-plant experiment did not find is that the response reliably buys you tolerance you can bank on.
So: believe the answer, discard the reasoning. Your plant is probably fine tonight because 4 °C is nowhere near freezing, not because it has toughened up over the summer.
So what does the cold actually cost you?
Time, mostly.
In that same study, CBD concentration and dry weight both went up as the temperature went up — but the authors are careful about why, and so should we be. Their reading is that this reflects the plants at warmer temperatures simply being further along by the time everything was harvested, rather than the cold having done damage. Which is the useful version for you: the cold plant is not ruined, it is behind.
That is consistent with the other thing we know about this plant. Cannabis photosynthesis has been measured to run best somewhere around 25 to 30 °C. A plant at 5 °C is nowhere near that, so it is doing very little work overnight. Autumn is not damaging your plant so much as it is putting it on short hours.
There is a name for the band you are in. In the general plant-physiology literature, warm-climate species are described as chilling sensitive roughly between freezing and about 12 °C — a zone where no ice forms anywhere, but photosynthesis is impaired and the plant runs poorly. That work is not about cannabis, and I am citing it for the mechanism rather than for a number that applies to your plant. But it is the right mental model: below about 12 °C you are in slow, and it is only at and below 0 °C that you enter damaged, which is a different process entirely — ice forming outside the cells and pulling water out of them until membranes fail.
And if you are worried that a cool finish is costing you potency, the evidence runs the other way. The most direct comparison of flowering-phase temperatures in medical cannabis found that the warmer regime reduced specialised-metabolite yield. Heat is the better-documented threat to what is in the flower; cold is a schedule problem.
The practical consequence is one most people have backwards. A cold snap is an argument for the harvest being further away, not closer. If your plant was not ready on Monday, a cold week does not make it ready by Friday.
The other thing cold reliably does is turn on the purple. In that same experiment, anthocyanin — the purple pigment — peaked at constant 8 and 15 °C, and was lower at both 0.5 and 22 °C. So autumn colour is a real temperature response and not a sign of harm. We have a whole page on telling that apart from the things that do mean harm.
The thing that actually ends outdoor grows in autumn is bud rot, not the cold — so “chop now or wait?” is a wetness question
Bud rot, and the season is the reason.
The most recent epidemiology of Botrytis on cannabis reports the disease peaking in September and October. That work was done in greenhouse-cultivated crops rather than in a field — worth saying plainly — but the crops were tracking the same seasonal swing in outdoor temperature and humidity that you are standing in. It is not a coincidence of the calendar: it is what cool, damp, shortening days do. Cold air holds less water vapour, so as the night temperature falls toward the dew point you get condensation: dew on the canopy, and hours of wetness that a July night would not have given you.
There is an irony worth absorbing, and I will flag it as my reading rather than a published comparison, because nobody has run that comparison. The temperatures at which this fungus infects are mild ones — the standard infection model, built in grapes, spans roughly 12 to 30 °C. That is autumn, not winter. So the nights you are afraid of are generally not cold enough to hurt the plant, while sitting squarely in the range that suits the thing that will.
And a mature bud is worse off than the air around it suggests. That same work measured the inside of a dense inflorescence running about 15 % higher in humidity and 2.5 °C warmer than the ambient air. Your hygrometer on the fence post is not reporting from where the fungus lives.
How much wetness it takes is the part worth internalising. In grapes — a different crop, and I am citing it only for the mechanism — Botrytis infection of berries was established after about four hours of wetness across a broad temperature range, with the risk climbing the longer the wetness lasted. Four hours is not a storm. Four hours is an ordinary clear autumn night with heavy dew.
So the correct anxiety, if you want one, is not about the forecast low. It is about how many consecutive hours your buds are wet, how dense they are, and whether anything has already chewed a hole in them. That is a longer conversation and it has its own page, including what to do after rain and how to price chopping early.
Why your porch thermometer is the wrong instrument
Here is the part that connects the two halves, and it is the reason a “the forecast says 2 °C, I am fine” conclusion can still go wrong.
On exactly the nights that produce a radiation frost — clear, still — surfaces do not sit at air temperature. They radiate heat to the open sky and end up colder than the air around them. This is why you can walk out to frost on the grass and on car roofs while the weather station is reporting a couple of degrees above freezing. As one instrument maker puts it in their technical notes, leaf and bud temperature can drop below freezing while air temperature stays slightly above it, and air temperature is therefore misleading, because the conditions that cause the frost are the same ones that pull plant temperature below air temperature — and not always by the same amount.
I am deliberately not going to give you a number for that gap. I looked for one in the authoritative sources and they do not publish one, which is itself the finding: the difference depends on the sky, the wind, the canopy and the night, so anyone handing you a fixed offset is guessing.
The usable version is a rule of thumb about which nights, not how many degrees. Clear and still is when the gap opens. Cloud or a breeze largely closes it.
The number nobody has published, and I would rather say so
There is real cannabis cold research — the two studies above are it, and there is more on the molecular side. What does not exist, as far as I can find, is a published kill temperature: no peer-reviewed “tissue dies below X °C for Y hours” for this plant, the way there is for commercial fruit crops, where whole extension programmes exist because the money justified the research.
You will nonetheless find a confident figure on grow sites, usually somewhere around −2 °C. I am not going to repeat it in a citation-shaped sentence, because I could not trace it to a measurement. It may well be sound field experience. Experience written in the register of a specification is still not a specification, and the way that number gets laundered into fact is by pages like this one quoting it.
What the absence means practically: treat a forecast at or below freezing as a real risk to manage rather than a quantity to calculate, and treat a forecast in the 2 to 10 °C range as a slow week rather than an emergency.
How to check this yourself
None of this needs anything bought.
Read the sky, not just the number. The night before, check whether the forecast is clear and calm or cloudy and breezy. Clear and calm with a low dew point is the frost setup. That one distinction will do more for you than a better thermometer.
Put a cheap thermometer at canopy height, where the buds are — not on the house wall, which leaks heat all night and will read warm. Extension guidance on frost monitoring is consistent on this point: the reading you want is from the height of the crop.
Go out at dawn, before the sun reaches the plants, and look. This is the measurement. Frost on the leaf surfaces, or leaves that are limp, dark and water-soaked in a way they were not the evening before, is actual freeze damage. Leaves that look normal at first light mean the night did not do what you feared, whatever the forecast said.
Then feel the buds. If they are wet at dawn, note roughly how long it takes them to dry once the sun is on them. Hours of wetness is the number that matters, and you can count it without instruments.
And judge ripeness separately, on its own evidence. A useful independent prior from the reference text for the field: modern cultivars generally mature 45 to 70 days after flowering is induced. If the forecast is telling you to chop at day 38, the forecast is not the reason you would be chopping. Our page on trichome colour covers why that call is harder than the loupe ritual implies.
What would show that we have this wrong
If you get a genuine hard freeze — several hours below freezing, not a brush with it — and the plants come through looking untouched, then the boundary I have drawn between “cold” and “freeze” is in a different place than I have put it, at least for your cultivar.
Equally, if you go through a cold, dry autumn and still lose buds to rot, then the wetness-hours mechanism is not doing the work I have credited it with in your conditions, and something else is driving it.
Either of those would be worth knowing. The claims on this page are specific enough to be wrong, which is the only kind worth publishing.
Where our device fits, stated narrowly
I should be straight about the limits here, because this is an outdoor question and what I build is a tent instrument. If your plants are in the back garden, the honest answer is that my product is not what is going to get you through October. A thermometer at canopy height and a look at the sky will.
What it does do is measure the number this page is about. It reads leaf-surface temperature directly and continuously, alongside air temperature, which means the gap between them is something it records rather than something you have to assume. That gap is the whole of the radiation-frost problem, and it is also the thing an air-temperature sensor cannot see by construction.
It does not predict frost, warn you about frost, or tell you to harvest. It keeps a dated record of what the plant actually experienced, so that when something does go wrong you are reading a log instead of reconstructing a week from memory. That is a narrower claim than the category usually makes, and I would rather make the narrow one.
Sources
- Kim, S.R., Basnet, P., Kovaleski, A.P. & Ellison, S.L. (2025). Anthocyanin accumulation, inflorescence dry weight and total cannabidiol content have different temperature optima in Cannabis sativa. Journal of Cannabis Research. DOI 10.1186/s42238-025-00311-w — flowering carried to harvest at constant temperatures from 0.5 to 22 °C; anthocyanin peaking at 8 and 15 °C and reduced at both ends; CBD and dry weight rising with temperature, which the authors attribute to plant maturity rather than to a temperature stimulus. One day-neutral type III cultivar, in growth chambers.
- Galic, A., Grab, H., Kaczmar, N., Maser, K., Miller, W.B. & Smart, L.B. (2022). Effects of Cold Temperature and Acclimation on Cold Tolerance and Cannabinoid Profiles of Cannabis sativa L. (Hemp). Horticulturae 8(6):531. DOI 10.3390/horticulturae8060531 — cultivars ‘FINOLA’ and ‘AutoCBD’; few effects of sequential cold treatments; those effects not moderated by cold acclimation, which tended to be negative across many responses, including decreased total CBD % and THC % against non-acclimated plants. This is the source for the acclimation section. The specific treatment temperatures and durations are not quoted on this page, because we worked from the published abstract and record rather than the full text.
- Mayer, B.F. et al. (2015). Cold acclimation induces distinctive changes in the chromatin state and transcript levels of COR genes in Cannabis sativa. Physiologia Plantarum 155(3):281–295. DOI 10.1111/ppl.12318 — cited only for the existence of a real molecular cold-acclimation response in this species, which is what makes the whole-plant result above worth stating carefully rather than flatly.
- Lyons, J.M. (1973). Chilling Injury in Plants. Annual Review of Plant Physiology 24:445–466. DOI 10.1146/annurev.pp.24.060173.002305, and Allen, D.J. & Ort, D.R. (2001). Impacts of chilling temperatures on photosynthesis in warm-climate plants. Trends in Plant Science 6(1):36–42. DOI 10.1016/S1360-1385(00)01808-2 — the chilling-sensitive band between freezing and roughly 12 °C, and the photosynthetic impairment in it. General plant physiology, not cannabis, cited for the mechanism only.
- Pearce, R.S. (2001). Plant Freezing and Damage. Annals of Botany 87(4):417–424. DOI 10.1006/anbo.2000.1352 — extracellular ice drawing water out of cells as the damage mechanism at and below freezing. Again general, not cannabis.
- Holweg, M.M.F.A. et al. (2025). High air temperature reduces plant specialized metabolite yield in medical cannabis, and has genotype-specific effects on inflorescence dry matter production. Environmental and Experimental Botany 230:106085. DOI 10.1016/j.envexpbot.2025.106085 — the direction of the temperature effect on what ends up in the flower. Cited for that direction only.
- Buirs, L., Lung, S. & Punja, Z.K. (2025). The epidemiology and management of Botrytis cinerea causing bud rot on greenhouse cultivated cannabis. Canadian Journal of Plant Pathology 47(5):443–463. DOI 10.1080/07060661.2025.2478250 — disease peaking in September and October; inflorescence interior about 15 % higher in humidity and 2.5 °C warmer than ambient. Greenhouse-cultivated crops, tracking the same seasonal pattern an outdoor grower sees, not a field study.
- Broome, J.C., English, J.T., Marois, J.J., Latorre, B.A. & Aviles, J.C. (1995). Development of an Infection Model for Botrytis Bunch Rot of Grapes Based on Wetness Duration and Temperature. Phytopathology 85(1):97–102. DOI 10.1094/Phyto-85-97 — infection after about four hours of wetness, rising with duration. A grape result, used here for the wetness mechanism only.
- Chandra, S., Lata, H., Khan, I.A. & ElSohly, M.A. (2008). DOI 10.1007/s12298-008-0027-x, and (2011). DOI 10.1007/s12298-011-0068-4 — cannabis photosynthesis temperature optima, around 25–30 °C.
- Caplan, D., Matzneller, P. & Gutierrez, G. Harvest and Post-Harvest, in Handbook of Cannabis Production in Controlled Environments (CRC Press). DOI 10.1201/9781003150442-10, §10.3.1 — 45 to 70 days after flowering induction for modern cultivars.
- Snyder, R.L. & de Melo-Abreu, J.P. (2005). Frost Protection: fundamentals, practice and economics, Volume 1. FAO Environment and Natural Resources Series 10 — the radiation-versus-advection distinction, and the clear-sky, calm-wind, inversion and low-dew-point conditions. Checked directly: it does not publish a leaf-minus-air figure, and neither does Penn State Extension’s frost guidance, which is why no such number appears above.
- Apogee Instruments, leaf and bud temperature sensor technical notes — leaf and bud temperature dropping below freezing while air temperature remains slightly above it, and air temperature being misleading because the conditions causing radiation frost also pull crop temperature below air temperature, “but not always by the same amount.” This is a sensor manufacturer’s technical page, not peer-reviewed literature, and is labelled that way deliberately.
- The absence of a published cannabis kill temperature or LT50 is a negative result from our own search, reported as such. So is the absence of any study directly comparing outdoor cannabis losses to bud rot against losses to frost — the framing that the autumn nights growers fear sit in the fungus’s range rather than the plant’s damage range is our reading of two literatures side by side, not a published comparison. The dawn inspection, the wetness-hours counting, the clear-and-calm rule of thumb, and the reading that a cold snap moves harvest further away rather than closer are first-party practice and judgement — not findings from a paper.
More of these at Answers. The device is at croplock.com.
Also read: Bud rot: how humid, for how long? · Can trichome colour tell me when to harvest? · Why is it purple?