Geothermal Driveway Heating: What It Really Takes to Melt Snow With the Ground

Ask whether geothermal can melt a driveway and you will usually get a confident yes, followed by a photo of a bare, steaming street in a snowstorm. The yes is real. The photo is usually of a system you cannot have, because it runs on naturally hot water from a well, not on a heat pump. The version a homeowner with a ground loop could build has to manufacture every degree of that heat. It is done — a state-funded bridge retrofit in Texas has been running and instrumented, at a lower supply temperature than the design manuals call typical. And if you idle the slab through the winter, the number that dominates your electric bill is not the melting load at all. It is what the slab costs to keep warm on the days it is not snowing.

Two very different things are called "geothermal snow melting"

The phrase covers two systems that share a slab and almost nothing else.

Direct-use geothermal. A well taps a naturally hot resource. The hot water passes through a heat exchanger, warms an antifreeze loop, and that loop runs through pipe in the pavement. Nobody manufactures heat; the ground supplies it already hot enough for the job. This is what the celebrated systems are, and it requires a hot resource under your feet. Almost nobody has one.

A ground-source heat pump. The loop under a typical home sits at ordinary earth temperatures, and the heat pump lifts that heat to whatever temperature the slab needs. Every degree of lift shows up as a lower coefficient of performance, the ratio of heat delivered to electricity consumed.

A third arrangement is worth knowing: a ground loop with no heat pump, circulating fluid at raw ground temperature through the slab. A 1969 research installation in Trenton, New Jersey did exactly that, and it is the only pavement system in these sources built to run with no compressor in between.

When a source says "geothermal" without saying which of these it means, check which one it means — the headline performance figures in the literature reviewed here come from the first kind.

The famous systems run on heat nobody had to make

The oldest geothermal pavement snow-melting system in the United States went into Klamath Falls, Oregon in 1948, built by the Oregon Highway Department: 450 feet of Esplanade Street on an 8 percent grade, 3/4-inch iron pipes 3 inches below the surface on 18-inch centers. In John Lund's account for the Geo-Heat Center at Oregon Institute of Technology, "the grid system was connected to a geothermal well with the heat transferred through a downhole heat exchanger to a 50-50 ethylene glycol-water solution that circulated at 50 gpm."

The well was naturally 143 °F. Supply to the grid ran from 100 to 130 °F, "the temperature drop in the grid was approximately 30 to 35oF," and the system "could provide a relative snow free pavement at an outside temperature of -10oF and a snowfall up to three inches per hour, at a heat requirement of 41 Btu/ft2/hr." The heat came out of the ground already hot.

The resource declined to 98 °F and the well was rehabilitated in 1992. In 1997, after roughly 50 years, the iron pipe failed from external corrosion. It was rebuilt in PEX, and "the entire cost of the reconstruction project was approximately $430,000" — a whole-project figure covering pavement and deck work on a 450-foot public street, not the price of a snow-melting system and not a driveway.

If you do not have a 143 °F well, you are in the second category, and that is the case the rest of this covers.

How much heat a driveway actually needs

Snow-melting slabs are sized in Btu per hour per square foot of surface. The current reference is Chapter 52, "Snow Melting and Freeze Protection," of the 2023 ASHRAE Handbook, HVAC Applications, reproduced by the Plastics Pipe Institute in its Recommendation J.

ASHRAE's Table 1 gives heat fluxes for 46 major US cities under the heading "Heat Fluxes Not Exceeded During Indicated Percentage of Snowfall Hours from 1982 to 1993, Btu/h-ft2," at percentiles of 75, 90, 95, 98, 99 and 100. Designing at the 75 percent column means the calculated load is met in three of every four snowfall hours in that record; the 100 percent column covers every hour in those twelve winters. Read those as steady-state figures, because that is what they are: ASHRAE warns that "to keep slab surfaces clear from snow during the first hour of the snowstorm, when the system is just starting to operate, heat fluxes up to five times greater than those indicated by steady-state analysis could be required." Each city has three rows for the snow-free area ratio, Ar:

• "Ar = 1.0 Snow-Free Area of 100% (i.e., no accumulation during snowfall)"
• "Ar = 0.5 Snow-Free Area of 50% (i.e., some accumulation allowed during snowfall)"
• "Ar = 0 Snow-Free Area of 0% (i.e., the surface is allowed to be covered with snow during heavy snowfall, melting snow from the bottom of the layer)"

PPI reproduces one city, Albany, New York, in Btu/h·ft²:

These are far higher fluxes than an indoor radiant floor is asked to deliver. The tubing-in-concrete idea is the same, but the load is an order of magnitude apart.

Two caveats travel with the numbers, and PPI states them itself: "The data in Table 1 ... was generated from a large dataset covering a period from 1982 -1993; this data may have shifted since that time. Additionally, those numbers are for only the surface of the slab, and specifically for a slab of width 20 ft (6.1 m) and an emittance value of 0.9."

Heat lost out the bottom and sides is on top of them. ASHRAE's chapter: "the values in Table 1 do not account for heat losses from the back and edges of the slab. Adlam (1950) demonstrated that these back and edge losses may vary from 4 to 50%, depending on factors such as slab construction, operating temperature, ground temperature, and back and edge insulation and exposure." For its own slab constructions, with ground at 4.5 °C at 600 mm depth, ASHRAE puts steady-state back losses at approximately 20 percent, and cites Spitler et al. (2002) for transient back losses of "approximately 12 and 29%, depending on storm conditions."

The load is a design choice, not a fixed number

The same city and the same twelve winters produce a design load anywhere from 37 to 321 Btu/h·ft², and the top figure is nearly nine times the bottom one. Nothing about the weather changed between those cells. What changed is what the owner decided to demand of the slab.

PPI puts it directly: "Designing for a snow-free area of 100% (Ar = 1.0) will require a much larger heat source, larger circulating pump(s), larger pipes, and other mechanical equipment... Designing for Ar values of 0.5 or 0 can dramatically reduce the size of this equipment, as well as energy consumption and operating costs." Its own residential example lands at the bottom of the table: "A residential driveway for a home in a neighborhood does not need to be 100% snow free in the worst case expected blizzard... the customer agrees to design parameters of Ar = 0 and Frequency distribution of 75%."

This is the lever that can bring the job within reach of a heat pump. The first decision is how much snow you are willing to look at.

What a ground-source heat pump can deliver, and what it costs in COP

That raises three questions: whether a heat pump is a legitimate heat source here, how hot it can go, and what the lift costs.

ASHRAE settles the first. The chapter's operating-cost equation has a term defined as "η b = combustion efficiency of boiler (or COP of a heat pump in heating mode), dimensionless." A heat pump appears in the standard's own cost equation alongside a boiler, and the number the method asks for is exactly the COP. ASHRAE adds that "other low-temperature waste, or alternative energy resources may also be used with or without heat pumps or heat pipes."

On how hot, WaterFurnace's catalog for its 5 Series 502W12 is plain: "Typically, water source heat pumps are limited to producing temperatures around 130°F which can make it difficult to retrofit an existing home that uses hot water for heat." The 502W12 is "capable of producing water temperatures up to 150°F," using R-134a rather than R-410A. Viega's hydronic design manual states: "Fluid supply temperature of 130°F is typical for snow melting applications." So the typical snowmelt supply sits exactly at the conventional water-source heat pump ceiling.

The concrete has its own ceiling. ASHRAE, in the SI edition of its chapter: "an upper limit to heat flux is typically bounded by the maximum allowable flow temperature for a pipe embedded in a thermal mass. This is particularly important for concrete, where entering fluid temperatures should not exceed 66°C." That figure is quoted as printed, in Celsius; the heat pump limits above are in Fahrenheit. A designer sizes to whichever limit binds first.

On the cost of that lift, the catalog is explicit: "When the system produces 130°F water from a 30°F earth loop, the ΔT is 100°F, and the COP is approximately 2.5. If the system is producing water at 90°F, the ΔT is 60°F and the COP rises to about 3.8, an increase of over 50%." Same hardware, same loop, a 50-plus percent difference in heat per unit of electricity, decided by the supply temperature you ask for. Melting from the bottom of a snow layer at a modest rate is a lower-temperature job than keeping a surface bare in a blizzard, which is another reason the Ar = 0 philosophy matters.

A snow-melting slab is a hot-water load, served by water-to-water geothermal heat pumps rather than the forced-air kind, and the slab side is a close cousin of geothermal radiant floor heating. At least one manufacturer names the duty. ClimateMaster's water-to-water product page states: "ClimateMaster's Tranquility TMW can be used for radiant floor heating, snow/ice melt...." with a "Load leaving temperature range from 25 to 140°F, -4.4 to 60°C (see submittal for specific model range)" and a "Source entering temperature range from 20 – 130°F, -6.7 – 54.4°C (see submittal for specific model range)." Snow and ice melt is a listed application for a geothermal water-to-water unit, with a leaving range that spans the typical 130 °F supply.

A working geothermal heat pump deicing system, documented by DOE

The sourced record includes a ground-source heat pump system that keeps pavement clear in service. It is a bridge, not a driveway, but it answers a question no catalog does: what temperature does a geothermal heat pump actually deliver to a slab when someone builds one?

The Texas Department of Transportation funded a deicing retrofit on a bridge in Tarrant County, adjacent to Fort Worth, studied by the University of Texas at Arlington. The Department of Energy's case study calls it "the first demonstration of retrofitting an existing bridge with a geothermal heat pump for deicing." The build: 16 boreholes, each 300 feet deep, and 4,300 square feet of deck heated with insulated hydronic loop panels on pipe loops spaced 6 inches apart.

Two cautions, because the DOE summary and the underlying TxDOT research report do not agree. DOE places the bridge on U.S. Highway 287 in Dallas; the research report puts it on SH 180 at West Village Creek relief in Tarrant County. DOE says the borehole field supplies "minimum 68 °F fluid to the heat pumps"; the research report measures the ground loop at 55 °F entering and 57 °F leaving, and uses 68 °F as a design ground temperature against a measured subsurface temperature of 69.7 °F — ground temperature and delivered fluid temperature are not the same quantity. DOE also describes "a thermostat that turns the system on automatically," while the report records that during testing "the whole system, including the four circulating pumps, was turned on/off manually." Where the two disagree, the figures here follow the research report.

Sixteen boreholes at 300 feet is 4,800 feet of borehole serving 4,300 square feet of deck. That is a scale marker for anyone imagining a driveway hung off a house loop.

Viega calls 130 °F typical for snow melting. WaterFurnace says conventional water-source heat pumps top out around 130 °F. This bridge ran well below both: the research report records a heat pump output temperature to the deck of 105 °F, notes that the maximum setting temperature allowed is 105 °F, and measured an average inlet to the bridge span of 90.8 °F, leaving at 81.5 °F. Neither source explains the choice, and nothing here should be read as an explanation. Set that against the catalog: about 2.5 at 130 °F and about 3.8 at 90 °F from a 30 °F loop. A bridge deck in north Texas is not a driveway in Albany. The bridge establishes something narrow: a geothermal heat pump deicing pavement is built, running, and documented by a federal agency, at a supply temperature well under the industry's typical figure.

Idling the slab all winter is the real energy bill

A snow-melting slab can sit cold until snow is detected, or be held just above freezing all winter, "idling," so it never has to warm up from a standing start. Idling sounds sensible. ASHRAE's own operating data say it is where almost all the energy goes.

Table 3 of the ASHRAE chapter, SI edition, gives "Annual Operating Data at 99% Satisfaction Level of Heat Flux Requirement" as "Annual Energy Requirement per Unit Area at Steady-State Conditions, kWh/m2." For Albany, New York, with 156 snowfall hours a year:

Melting energy falls by more than half as the design choice relaxes, 32.0 to 13.8. Idling energy barely moves, 344.5 to 342.0, and in every row it is roughly ten to twenty-five times the melting figure. The design choice that dominates equipment size has almost no effect on annual energy, because annual energy is dominated by keeping the slab warm on the days it is not snowing.

The pattern holds across the cities ASHRAE tabulates, in kWh/m² per year:

Cold winters with modest snowfall are the worst case, because the slab is held above freezing for months to be ready for a few storms.

ASHRAE draws the conclusion itself: "However, idling the slab during the entire winter, as given in Table 3, may be too costly, and unnecessary if the main purpose is to reduce high snow-melting surface heat flux at start-up. For example, the annual energy requirement for idling is 45 times more than that for snow-melting in Chicago, A r = 0.5. Therefore, a cost-effective operation may require starting the system to idle only before an anticipated snow."

One caveat on that "45 times," because the chapter is inconsistent with its own table. Table 3's Chicago row gives 17.0 kWh/m² melting against 355.7 idling at Aᵣ = 0.5, a ratio of about 21. The 45-to-1 ratio appears in the Chicago row at Aᵣ = 0 (7.1 against 316.7). The direction of ASHRAE's point is unaffected and its own table supports a 21-to-45-fold gap depending on the design choice, but the specific pairing of "45 times" with Aᵣ = 0.5 does not reconcile, and it is worth knowing before you quote the number at a contractor.

The whole case for a ground-source heat pump is its COP. Pair it with all-winter idling and you have an expensive machine running the most expensive strategy in the table, holding a slab warm for several hundred kilowatt-hours per square metre a year to deliver a few tens. A good COP does not rescue that; the control strategy is what moves the number: idle only ahead of an anticipated storm, or not at all, and the annual energy collapses toward the melting column.

The figures carry a precondition. ASHRAE: "Idling and melting data are based on slab surface temperature control at 0°C during idling, which requires a slab temperature sensor. Without a slab temperature sensor, operating costs will be substantially higher." On the hardware: "Snow detectors located in the heated area activate the snow-melting system when precipitation (snow) occurs at a temperature below the preset slab temperature (usually 4°C)," and "a remote temperature sensor is commonly installed midway between two pipes or cables in the slab; the set point is adjusted between 5 and 15°C." A snow detector and a slab sensor are what let a system sit in the melting column instead of the idling one.

The ground loop with no heat pump: Trenton, 1969

The historical case differs from the bridge in one decisive way: no heat pump. Lund describes a 1969 research installation in Trenton, New Jersey that circulated "an ethylene glycol-water mixture between pipes embedded 2 inches below the pavement surface and a horizontal grid buried 3 to 13 feet below the pavement on 2-foot levels. The total length of the ground pipes was twice as long as the pipes in the pavement." The only powered component was the circulation pump.

"The measured undisturbed ground temperature at 7 feet depth varied between 48 and 57oF during the winter and the antifreeze temperature ranged between 40 and 52oF during most of the snow storms." The result: "Typical measured snow melting rates were 1/4 and 1/2 inches per hour when the corresponding air temperature ranged between 20 and 35oF." Klamath Falls, on 100 to 130 °F direct-use water, handled up to 3 inches per hour at -10 °F. That is the gap raw ground heat leaves, and a heat pump's job is to close some of it at the COP cost above.

Trenton still performed: "The performance of this ground system proved to be superior to that of a companion 68 Btu/h/ft2 electric pavement heating system while requiring only about 2% of the electrical power to operate the circulation pump." The catch: "One of the draw-backs with the system was the expensive excavation required for placement of the ground pipes."

For a homeowner planning a ground loop, that is worth raising early. The sizing rules in our geothermal loop pipe guide assume a building load, not a slab open to the sky, and a driveway belongs in the conversation before the excavator arrives, which is why it comes up in our overview of geothermal heat pump installation. Whether a loop sized for a house has anything left over for a driveway is not something the sources answer, and we will not guess.

Pipe, fluid, and the parts list

Pipe. Lund: "Present practice in the U.S. is to use plastic pipe with iron for the header pipe. Typical plastic pipes are of a cross-linked polyethylene (PEX), that according to ASTM standard F 876, can handle 180oF water at 100 psi or 200oF water at 80 psi." Iron in the slab is what corroded out of Klamath Falls after about 50 years; Lund notes "the corrosion rate approximately doubles for each 18oF rise in temperature."

Fluid. PPI describes closed-loop systems circulating "warm anti-freeze mixtures (e.g., glycol and water) through plastic piping." Uponor's snow and ice melting manual gives the rule: "for freeze protection, choose a glycol solution concentration that prevents the formation of ice crystals at a temperature of at least 5°F (3°C) colder than the lowest expected ambient temperature", and warns, "Do not use burst glycol concentrations to protect snow and ice melting systems." The same manual states that snow and ice melting systems are "typically designed at a 25°F ∆T. This is the basis for the design." The manual's fluid data are attributed to The Dow Chemical Company.

The rest. PPI's list for a closed-loop installation: it will "typically include a heat source, circulating pump, piping, manifolds, controls, and other mechanical devices such as compression tanks and safety valves (see local codes for safety requirements)." The heat source is where the direct-use versus heat pump distinction lives; everything else is the same regardless.

The older Class I / II / III framework

You will still see an older sizing method. Lund's paper uses it: "Chapman (1957) classifies snow melting installation according to type as Class I, II or III," with Class I covering "residential walks or driveways," and his design table comes from the 1995 ASHRAE Applications Handbook, giving New York City 121, 298 and 342 Btu/h·ft² for the three classes and Chicago 89, 165 and 350. The classification dates to 1957 and the figures to 1995. The current handbook and PPI both work in snow-free area ratio and frequency percentile instead, so a source still quoting "Class I" is describing the older approach.

What cost figures actually exist

Installed cost. The engineering literature prices public works, not driveways: no residential figure turned up, by heat pump or any other heat source. Consumer sites and vendor pages publish ranges; those are not engineering sources, and averaging them would produce a number with nothing behind it. What does exist is public-works pricing. The Geo-Heat Center reports that on the Wall Street Bridge project in Klamath Falls, "the entire geothermal portion of the snow melting project was awarded at $170,000, which figures out at $16.45 per square foot" — with the author's own caveat that "costs for state projects tend to be two to three times higher than private projects due to the requirements to pay prevailing wages and rigorous inspection." Read that $16.45 as a municipal, prevailing-wage number for a bridge deck, not a quote for your driveway.

Operating cost. No verified dollar figure exists for operating a heat-pump-driven snow-melting slab. A wattage-per-square-foot figure is often quoted for electric resistance systems, and it does not transfer to a hydronic system run by a heat pump. What does exist is ASHRAE's annual energy table above, in kWh/m² per year at the slab. That is the energy the slab needs, not the electricity a heat pump buys to supply it; ASHRAE's operating-cost equation brings in the heat source's efficiency, defined for a heat pump as its COP in heating mode, to get from one to the other. The dollar figure then depends on your rate, your slab area, your climate, your design choice, and above all whether you idle. The honest way to get one is to run ASHRAE's method with your own inputs.

What the sources do settle is which way each cost moves. PPI says designing for Ar = 0 or 0.5 "can dramatically reduce the size of this equipment, as well as energy consumption and operating costs." WaterFurnace's own performance table for the NHW084, at a 30 °F entering source temperature and 15 gpm source flow, prints a COP of 3.9 at a leaving load temperature of 88.8 °F against 2.5 at 128.4 °F — asking for cooler water buys back most of the efficiency. ASHRAE says all-winter idling can be 45 times the melting energy in Chicago at Ar = 0.5, and the remedy is to idle only before an anticipated snow, on a slab sensor. Trenton says the expensive part of a ground-coupled slab was the excavation. Those are the levers; the dollar figures on the other end of them come from a designer with your site in front of them.

The honest verdict

Geothermal driveway heating is a real application, a demanding one, and the residential record is thinner than the enthusiasm around it.

The systems that made the idea famous are direct-use installations: the heat comes from a naturally hot well, handed to the pavement loop through a heat exchanger rather than made by a machine. A heat pump is a recognized heat source in ASHRAE's method and can reach the typical 130 °F supply, at a COP of about 2.5 from a 30 °F loop against about 3.8 at 90 °F. It is done: DOE documents four geothermal heat pumps on 16 boreholes supplying 100 °F fluid to 4,300 square feet of Texas bridge deck, and ClimateMaster lists snow and ice melt as a duty for its Tranquility TMW. What the sourced record does not contain is a residential driveway on a ground-source heat pump with published performance.

Equipment size is set by how much snow you tolerate, 37 to 321 Btu/h·ft² in Albany. Annual energy is set by whether you idle, 32.0 against 344.5 kWh/m² a year in the same city. Pick Ar and percentile first, size for that flux plus back losses, stay under the concrete ceiling, decide in writing that the slab will not idle all winter, and only then ask what equipment can make that temperature from your loop and what it costs to run.

Frequently asked questions

Can my existing geothermal heat pump melt my driveway?

Not as a bolt-on, and not on anything the sources can promise. Whether your unit and loop have capacity beyond the house depends on a load calculation nobody has published for your site. The application itself is established: DOE documents a geothermal heat pump deicing system on a Texas bridge, and ClimateMaster lists snow and ice melt as a duty for its Tranquility TMW.

Should I keep the slab warm all winter so it is ready for snow?

ASHRAE's data say that is where nearly all the energy goes: for Albany, melting is 32.0 kWh/m² a year at Ar = 1 and idling is 344.5. The chapter's remedy is "starting the system to idle only before an anticipated snow," and its figures assume a slab temperature sensor; without one, operating costs "will be substantially higher."

Does a heated driveway have to stay completely bare?

No, and deciding that it does not is the biggest thing you can do to shrink the system. In Albany, full snow cover allowed during heavy snowfall at the 75 percent column is 37 Btu/h·ft², against 321 for a bare surface at the 100 percent column. PPI's own residential example designs at Ar = 0 and 75 percent.

How much does it cost to install or run?

There is no verified installed-cost or operating-cost dollar figure for a residential heat-pump snow-melting system in any source this site will cite. The one real installed number, approximately $430,000, is for rebuilding 450 feet of public street in Klamath Falls. Your figure has to come from a designer with your slab, your climate, and your Ar, percentile and idling decisions in front of them.