Takeaways
- The comparison is about source stability, not about efficiency in the abstract. Below the frost line, ground temperature is effectively constant year-round. Outdoor air is not, and it is coldest at the precise moment the building needs the most heat.
- The federal cold-climate specification tells you how much air-source performance is expected to fall. A qualifying cold-climate air-source heat pump must demonstrate a coefficient of performance of at least 1.75 at 5 degrees Fahrenheit with at least 70 percent capacity retention. A ground-source machine on a properly sized loop is working against roughly 50 degree fluid at the same moment.
- Air-source is dramatically cheaper to install and that will usually decide single-building retrofits. There is no drilling, no loop field, no easement, no site access problem.
- Ground-source wins on the margin, and margins compound at scale. The Vandalia community system was modelled at a collective network coefficient of performance of 5.5 on an ambient loop running 45 to 95 degrees Fahrenheit, a figure no air-source system reaches in an Illinois January.
- Do not compare on rated efficiency. Compare on design-day electrical demand. Auxiliary resistance heat is where air-source lifecycle cost is actually decided, and it does not appear in a nameplate rating.
The physics, stated once
A heat pump does not create heat. It moves heat from a colder place to a warmer place, and it spends electricity to do it. The coefficient of performance, universally abbreviated COP, is simply how many units of heat arrive in the building per unit of electricity consumed. A COP of 4 means four units of heat for one unit of electricity.
How much electricity that costs depends almost entirely on the temperature difference the machine has to bridge. Lift the heat a short distance and the COP is high. Lift it a long distance and the COP falls. This is not an engineering detail that better equipment eventually solves; it is thermodynamics.
So the design question reduces to one thing: what is the temperature of the source, on the day the building needs the most heat?
Air-source
The source is outdoor air. On a design winter day in a cold climate that might be 5 degrees Fahrenheit or below. The machine has to lift heat from 5 degrees to a delivery temperature suitable for the building. That is a large lift, at exactly the hour the heating load peaks. The two curves move against each other.
Ground-source
The source is the ground, reached through a fluid loop. Below the frost line, ground temperature converges on the local annual average air temperature and its seasonal swing damps out rapidly with depth, which is a straightforward consequence of the thermal diffusivity of soil and rock rather than a property of any particular product. The lift is smaller, and it is roughly the same lift in January as in March.
That is the whole argument, and everything below is a consequence of it.
How far apart, in numbers
Cold-climate air-source heat pump performance is not a matter of opinion, because the qualifying specification states it. To qualify as a cold-climate unit, equipment must reach at least 8.1 HSPF2, the heating seasonal performance factor measure, and must demonstrate a COP of at least 1.75 at 5 degrees Fahrenheit while retaining at least 70 percent of its capacity.
Read that as the floor, not the typical value, and note what it concedes. At the design condition, a qualifying cold-climate machine is permitted to deliver 1.75 units of heat per unit of electricity and to have lost 30 percent of its capacity. Both of those are enormous changes from its rated mild-weather performance.
| Attribute | Air-source | Ground-source |
|---|---|---|
| Heat source | Outdoor air | Ground, via fluid loop |
| Source temperature at winter design | Design outdoor air, commonly at or below 5°F in cold climates | Loop fluid, broadly stable year-round |
| Qualifying COP at 5°F | At least 1.75, with at least 70% capacity retention | Not applicable; performance is not set by outdoor air |
| Capacity at design condition | Reduced, by specification down to 70% of rating | Substantially unchanged |
| Auxiliary resistance heat | Commonly required at design conditions | Commonly not required |
| Installed capital cost | Substantially lower; no drilling | Substantially higher; borefield dominates |
| Site requirements | An outdoor pad and clearances | Drilling access and loop field area or depth |
| Vandalia network modelled COP | — | 5.5 collective, ambient loop 45 to 95°F |
Why the source temperature is the whole argument
The cost that hides in the electrical service
The most consequential difference between the two systems does not appear in a nameplate or an efficiency rating. It is auxiliary heat.
When an air-source heat pump loses capacity in cold weather, something has to make up the shortfall, and in most installations that something is electric resistance heat. Resistance heat has a COP of exactly 1. Every unit of electricity becomes one unit of heat, which is the definition of the thing a heat pump exists to avoid.
This has three consequences that a seasonal efficiency rating hides:
- The worst efficiency arrives on the highest-load day. Seasonal averages smooth this away. Utility bills do not, and neither do peak demand charges on a commercial tariff.
- The electrical service has to be sized for it. Resistance backup can drive the panel and service upgrade, and in a retrofit that upgrade is sometimes the single largest line item.
- It shows up in the utility's winter peak, not just the customer's bill. For a municipality or cooperative evaluating electrification across many buildings at once, the aggregate winter peak created by resistance backup is a system planning problem, not only a customer economics one.
A ground-source system working against a stable loop generally does not need meaningful resistance backup, because its capacity has not collapsed at the design condition. That is where the lifecycle difference is actually created.
A framework for deciding
Rather than a payback rule of thumb, the useful question is a sequence.
1. What is the design condition, and what does the load look like at it?
Start with a real load calculation at the actual design temperature for the site. Everything downstream depends on this and nothing substitutes for it.
2. What does air-source cost to run at that condition, including auxiliary heat?
Model the resistance backup explicitly as its own line, at COP 1. If the analysis does not have a separate auxiliary heat line, it is not comparing the two systems, it is comparing two brochures.
3. Does the site have a service constraint?
If an air-source retrofit triggers a service upgrade and a ground-source one does not, that upgrade cost belongs in the air-source column. It frequently closes a large part of the capital gap on its own.
4. How many buildings are there, and can they share a loop?
This is the question that changes the answer most. Borefield cost per building falls sharply when a loop serves many buildings with diverse load profiles, because the field is sized against the diversified peak rather than the sum of individual peaks. The Vandalia system illustrates this directly: 2,005 tons of diversified heating capacity, explicitly described as diversified, serving 241 homes, three schools, a hospital, commercial buildings and a municipal pool from one shared field.
5. What is the counterfactual fuel, and how has its delivered price behaved?
Covered separately in our analysis of gas price volatility. In short, the delivered residential gas price has behaved as a one-way ratchet even when the underlying commodity collapsed, which is a stronger argument for a fixed-cost thermal plant than any single-rate escalation assumption.
The honest summary
For a single existing house on a constrained lot, air-source is usually the right answer, and the reasons are capital cost and site access rather than any deficiency in the physics of ground-source.
Ground-source becomes compelling in three situations: where the design condition is severe enough that air-source capacity loss and resistance backup dominate the operating cost; where an electrical service upgrade would be triggered anyway; and above all where enough buildings can share one loop field that the drilling cost is spread across a diversified peak. The third of those is why community-scale thermal energy networks exist at all, and why the interesting engineering question is usually not which machine, but how many buildings.