With geo there is a much bigger risk factor related to installer competence than there is with air-source heat pumps. The system designers & installers will make or break the potential COP of the GSHP. While this is true for air source heat pumps too, the number of ways you can screw it up are fewer. (Though the more idiot-proof you make something, the more creative the idiots become.) With GSHP it can be fairly difficult to fully vet the contractor ahead of time, but as an engineer with geo experience you'll have a much better shot at it than most.
But 3 tons of load for a 2750' house implies a IRC 2009 code min (or worse) house, and if the house is not yet built there are numerous ways to cost effectively reduce those loads, often dramatically. (For reference, I live in a ~2700' antique 1.5 story 2x4 framed house in US climate zone 5A, and have design heat @ +5F that would be covered by almost any 3 ton system. Your new house should do better- a LOT better!) If you can bring the loads to under 2 tons a 1.5 ton Fujitsu 18RLFCD ducted mini-split (20K nominal output @ +17F) would be comparatively cheap to install- less than half that of a 3-ton variable speed ASHP, with comparable or better efficiency. (In some cases it will even have higher capacity at low temp than a 2-3 ton variable speed units from the aforementioned vendors.)
The ton or more of load that disappears has a lifecycle nearly as long as the house, and uses no power, so even if the building envelope upgrades with a 1.5 mini-split costs a bit more, it can still be "worth it" on energy savings alone a lifecycle basis. But the higher performance envelope done right will also be more comfortable and more resilient. With the possible exception of radiant floors, a higher thermal performance house buys more comfort than any heating system will. That higher thermal performance is far cheaper now (at the design stage) than it ever will be after the fact.
Whether building a code-min house or a higher performance house,
taking the extra time to define the primary air barrier layer during the
design phase, then enforcing/inspecting that during construction is
some of the cheapest performance you can ever buy. Most infiltration
& ventilation defaults in load calculation tools are on the
ridiculously high side- higher than would be found in a real IRC 2012
code house. Code max air leakage is 3 air exchanges at 50 pascals
(3ACH/50), but it doesn't take rocket science builders to build to half
that or less. It's possible that your 3 ton load numbers have an
excessive infiltration or ventilation loss driving the numbers beyond
reality here.
A typical IRC 2012 code min house comes in at a heat load/floor ratio of about 10-12 BTU/ft^2- hr. A pretty good higher-R house (without taking it to the PassiveHouse extreme, or even to Net Zero Energy), can come in at 6-8BTU/ft^2- hr without a huge expenditure. When looking at lifecycle financial rationality IRC 2012 code min isn't even close to blowing the budget- there's plenty of ways to bump it up cost-effectively. As a good starting point, take a peek at Table 2 p.10 of this document:
http://buildingscience.com/file/580...n=GouEIX9YNote- those are "whole assembly R", not center-cavity R: eg: a code-min 2x6/R20 wall is really R14-R15-ish after factoring in the thermal bridging of the framing. To hit R20 would require adding R5 continuous exterior sheathing (not a recommended stackup unless the continuous insulation is rigid rock-wool, due to dew-point/drying resilience concerns)
You are most likely in US climate zone 5, but possibly zone 6. In Table 2 they are suggesting that R30 whole-wall is in the mid-range of what might be financially rational on a full lifecycle basis, depending on what your future energy cost inflation projections and current material, labor, and financing costs are. An example of a ~R30-ish whole-wall stackup would be 2x6 / R20 + 3" of exterior rigid foam.
In cheap energy/high labor markets that may not be lifecycle-rational unless you are projecting high energy cost inflation at some point in the next 50 years. But a true R20-R25 whole-wall stackup is almost certainly going to be financially rational. A very resilient R25-ish wall in that range would be 2x6/R20 using damp-sprayed cellulose or open cell foam cavity fill, with a 1" layer of foil faced polyiso on the exterior of the sheathing (R5.5-ish derated for climate temperature and layer within the stackup) and 1" 1.5lb density "Type-II" EPS on the exterior of that (R4.5-ish temperature up-rated for climate and layer within the stackup.) If that wall is too fat for you, try 2x4/cellulose + 1.5" polyiso + 1-1.5" EPS, which is at worst an inch thicker than a 2x6 wall that has no insulating sheathing, but comes in at ~R22-R25 whole wall, compared to R14-R15 for a code-min house, reducing the wall-loss figures by about 1/3.
Bumping up to R65-R75 in the attic from a code-min R49 with another 4-6" blown cellulose is dirt cheap, but the extra depth required means it needs to be designed into the roof trusses or framing ahead of time to accommodate it.
It goes on...
By the time you've spent $15-20K on going beyond code min on a least-cost methods basis it's likely that you will have gotten the heat load under 24,000 BTU/hr, possibly under 20,000 BTU/hr, at which point the higher cost mechanical systems become less interesting.
Before finalizing any of the decisions it's worth downloading a copy of
BeOpt and simulating the energy use of the house playing "what if?" games on different envelope changes using budgetary numbers (or real quotes) on different options.