The room-by-room HVAC load calculator used to say "no basements or skylights" on its how-it-works page. That ruled out much of the northern US housing stock. Both now work. Before shipping the basement formula I checked it against an independent solve, and the floor half came out further off than I expected.
The formula
A basement wall loses heat through the soil to the ground surface. Heat leaving a deep strip of wall has a longer path to the surface than heat leaving the top foot. Latta and Boileau (1969) modelled those paths as circular arcs. In the closed form ASHRAE Fundamentals gives, a wall from grade down to depth z has an average U-factor of
U = 2k / (π z) × ln((z + 2kR/π) / (2kR/π))
where k is the soil's conductivity and R is the wall's own resistance. A basement floor of shortest width w gets a similar log form. Both work against the ground surface design temperature: the annual mean air temperature less its seasonal swing, not the winter design air temperature. For a bare 8 in. concrete wall 7 ft deep in moist soil (k = 0.8 Btu/h·ft·°F) that gives U = 0.177 Btu/h·ft²·°F, about twice an R-13 frame wall's.
The check
eval/basement-fd.mjs solves 2-D steady conduction in the soil around a long basement on a grid of 0.25 ft cells, with conjugate gradients, about 200 lines of plain JavaScript. The room is held at 1 and the ground surface at 0. The heat entering through the wall and floor matches the heat leaving the surface to ten decimal places. A second run holds the deep ground at the annual mean, which is warmer than the surface on a winter design day. The closed form ignores that.
| case | wall U, formula | wall U, soil model | floor U, formula | floor U, soil model |
|---|---|---|---|---|
| 24 ft wide, 7 ft deep, bare | 0.177 | 0.198 (formula 10% low) | 0.030 | 0.041 (27% low) |
| 24 ft wide, 2 ft deep, bare | 0.351 | 0.390 (10% low) | 0.049 | 0.060 (18% low) |
| R-10 on the wall | 0.058 | 0.056 (3% high) | 0.030 | 0.053 (43% low) |
| R-10 under the floor | 0.177 | 0.210 (16% low) | 0.023 | 0.024 (5% low) |
The wall formula holds up: about 10% low on bare concrete and within 5% once the wall is insulated. The floor formula is the weak half. It treats the floor as if the wall weren't there. In the soil model, heat from the floor's edge leaves through the soil beside the wall. When the wall is insulated, that soil stays colder, and the floor's loss rises by more than a quarter (0.041 to 0.053) while the formula doesn't move. With the deep ground held at its annual mean, the gap ranges from 31% low to 24% high.
Does it matter? On a worked cold-climate house (the calculator's example ranch moved to 44°N, -5 °F winter design, over an 1,800 ft² basement 6 ft below grade) the basement floor is about 2,350 Btu/h of a 53,000 Btu/h heating load. A 27% miss there is about 1% of the total. The page states the gap and doesn't patch it with a fudge factor.
The same house shows why basements had to be modelled at all. Its heating load is 53,021 Btu/h over bare basement walls against 42,033 on a slab. R-10 on the basement walls brings it to 45,844.
Skylights
Skylights take an area, a tilt and the direction the roof slopes. The sun on the glass is worked out hour by hour at that tilt, and the skylight's area comes out of the attic ceiling. On the example ranch two 8 ft² skylights add about 1,300 Btu/h at the house peak, a bit over a tenth of a ton, and move the peak from 3 pm to 2 pm. A flat skylight in July at 42°N gets more than twice the direct sun of the same glass in a south wall.
The method, the tests and the solver are in the source. As before, this is an estimate from an open method, not an ACCA Manual J report.