Ground-Source vs Air-Source Heat Pumps in Extreme Cold
Updated 8 October 20267 min readGeothermal & Heat Pumps
Both machines run the same refrigeration cycle. What separates them in a cold climate is timing: an air-source heat pump loses capacity precisely when the building needs the most heat, while a ground loop is still drawing from soil that has barely noticed winter. Everything else — defrost cycles, back-up heat, sizing — follows from that one divergence.
Key takeaways
- Capacity and demand move in opposite directions for an air-source machine: as the air cools, the building wants more heat and the machine can deliver less.
- The temperature where those two curves cross is the balance point, and it is the single most useful number in a cold-climate design.
- Ground loops sidestep the divergence rather than solving it: the source barely moves, so capacity on the coldest night is roughly what it was in October.
- Defrost is a real energy cost unique to air-source machines, and it is worst in the damp, near-freezing conditions that are common in maritime winters.
- Modern cold-climate air-source machines narrow the gap substantially with vapour injection and variable-speed compressors, without closing it.
On this page
Same cycle, different source
Nothing inside the two machines disagrees. Both evaporate a refrigerant at low pressure to collect heat, compress it, condense it indoors to release that heat, and expand it back — the cycle set out in how geothermal heat pumps work.
The difference is entirely in what they collect from. One is connected to the air, which in a cold climate can swing 40 °C (72 °F) across a year and 15 °C (27 °F) across a day. The other is connected to soil a few metres down, which moves by a few degrees across a whole season.
The divergence that defines a cold-climate design
Here is the awkward geometry of air-source heating. As the outdoor air gets colder, two things happen at once, and they pull in opposite directions.
The building wants more heat. Heat loss is roughly proportional to the difference between inside and outside, so demand rises in a straight line as the temperature falls.
The machine can deliver less. A colder evaporator means lower refrigerant pressure and density, so the compressor moves less refrigerant mass per revolution and the machine's heating capacity falls. Meanwhile the temperature lift has grown, so each unit of heat costs more electricity.
The temperature at which the two lines cross is the balance point. Above it the heat pump covers the load on its own. Below it there is a shortfall, and something has to fill it.
A ground-source machine does not escape the physics — its lift grows as the delivery temperature rises and its loop cools — but its source line is nearly flat, so its capacity line is nearly flat too. The crossing either happens far below any temperature the site sees, or never.
What the cold does to an air-source machine, step by step
- Lower suction pressure. Colder air means a colder evaporator and a lower-pressure, less dense refrigerant vapour. The compressor is a volume pump, so less mass flows per revolution.
- Higher compression ratio. The gap between evaporating and condensing pressure widens, which lowers compressor efficiency and raises discharge temperature.
- Lower COP. The lift has grown, so more electricity is required per unit of heat delivered — the relationship laid out in the COP table in the geothermal article.
- Defrost, in the wrong conditions. Between roughly −5 °C and +5 °C (23 °F to 41 °F) with damp air, moisture freezes onto the outdoor coil. Ice insulates it and blocks airflow, so the machine periodically reverses the cycle to melt it off — taking heat *from* the building, and pausing heating while it does.
Defrost is the reason a damp winter at 0 °C can be harder on an air-source machine than a dry one at −10 °C. A ground loop has no equivalent: there is nothing to frost.
Below the balance point: what fills the gap
Three answers, and most cold-climate systems use a combination.
- Electric resistance elements. Simple, cheap to fit, and with a coefficient of performance of exactly 1 by definition. Every kilowatt-hour they consume delivers one kilowatt-hour of heat, exactly as a solar diversion load does, against three or four from the heat pump — which is why a system that quietly runs on resistance heat for weeks looks like a heat pump that has failed, even though it is still warm inside.
- A second heat source. A boiler or stove sized for the extremes, with the heat pump doing the bulk of the season's work.
- A bigger heat pump. Moves the balance point lower, at the cost of a machine oversized for the mild majority of the season, which then short-cycles when lightly loaded.
The design question is not "will it need back-up?" but "how many hours a year will back-up run, and how much of the season's energy is that?" — the same split between peak power and energy that structures an off-grid load calculation. A machine that meets 95% of the annual load and leans on an element for a handful of nights is generally a better design than one sized for the worst hour. The trade is sharpest in an off-grid solar system, where the coldest weeks bring both the weakest sun and the heaviest heating load.
What cold-climate machines changed
Air-source performance in deep cold is not what it was, and two developments did most of the work.
Vapour injection. An extra port part-way through compression admits a second, intermediate-pressure stream of refrigerant tapped through an economiser. This cools the compressor internally, allows a higher compression ratio without excessive discharge temperature, and raises the mass flow reaching the condenser — which is precisely the quantity the cold was taking away.
Variable-speed compressors. Rather than one fixed output switched on and off, the compressor modulates. In mild weather it runs slowly and continuously — more efficient than short cycling — and in deep cold it runs hard, so nameplate capacity is available where it is needed rather than averaged away.
The result is machines rated with real capacity at low outdoor temperatures. The gap to ground-source narrows; it does not vanish, because the source is still air.
The questions that actually decide it
| Air-source | Ground-source | |
|---|---|---|
| Source temperature on the coldest night | The coldest air of the year | Within a few degrees of the annual average |
| Capacity when demand peaks | At its lowest | Essentially unchanged |
| Defrost losses | Real, worst in damp weather near freezing | None |
| Back-up heat | Usually required below the balance point | Often not required at all |
| Installation | Outdoor unit; no ground works | Trenches or boreholes, and the land or access for them |
| Noise outdoors | A fan and compressor running in winter | Nothing outside |
| What can go wrong slowly | Coil fouling, defrost control faults | An undersized loop cooling the ground year after year |
Behavioural comparison, not a recommendation. The numbers that settle a specific case — capacity at the design temperature, balance point, seasonal efficiency — come from manufacturer data for the machine and from a heat loss calculation for the building.
Three questions decide most real cases. How cold is the design temperature, and for how many hours a year? What delivery temperature does the building need — because a low-temperature emitter system flatters both machines and rescues neither from a badly chosen balance point? And is there ground to work with, in area, access and geology?
Where the answers point to ground-source, the follow-up is which loop geometry suits the site, which is the subject of closed-loop versus open-loop systems. Where they point to air-source, the thing worth understanding before the first cold snap is what normal defrost behaviour looks like, so that it is not mistaken for a machine freezing up.
Frequently asked questions
At what temperature does an air-source heat pump stop working?
It does not stop; it shrinks. Capacity and coefficient of performance decline steadily as outdoor air cools, and each machine has a published minimum operating temperature below which it will not run. The practical limit is usually economic and comfort-related — the point where back-up heat is doing most of the work — rather than a cliff.
Does a ground-source heat pump lose capacity in winter at all?
A little. The loop cools somewhat during the heating season as heat is withdrawn faster than the surrounding ground replaces it, so the source temperature drifts down over the winter and recovers in summer. The swing is small compared with air temperature, which is the entire point.
Is back-up resistance heat a design failure?
No, if it is sized and controlled deliberately. Sizing a heat pump for the single coldest hour means oversizing it for the other nine hundred hours of the season, which brings its own inefficiency. A modest element covering the extremes is often the better engineering choice.
Do heat pumps work in very cold climates in practice?
Yes, and increasingly they are the normal choice, but the design margins get tighter. Cold-climate air-source machines are rated at low outdoor temperatures specifically so this can be checked, and ground-source removes the question entirely at the cost of excavation.
Which is better if I can have either?
Ground-source, on thermodynamics alone, in a cold climate. Whether that advantage is worth the ground works depends on the site, the heating season length and how low the design temperature is — which is a question about the building and the ground, not about the machines.
Sources
Named organisations whose published material underpins this article. Where no link is given, the source is named rather than linked.
- U.S. Department of Energy, Geothermal Technologies OfficeReference material on ground-source and air-source heat pump operation.
- International Energy Agency (IEA)Technology overviews covering heat pumps in cold climates.
- Manufacturer performance dataCapacity and COP at low outdoor temperatures, minimum operating temperature and defrost behaviour are product-specific and published per model.
Editorial Team
Research, drafting and review
Articles are drafted from primary engineering and physics references with AI-assisted tools, then reviewed and fact-checked line by line by a human editor before publication. We publish explanations, not recommendations: no products, no pricing, no country-specific rules, and no invented author personas.
Last reviewed 8 October 2026. How we research and review