A passive house with rooftop solar reduces this home's 30-year carbon by 74%.

The 5-digit ZIP code resolves through a static crosswalk (39,366 ZIPs, 54 states/territories) to a 2-letter state code. State then drives all regional parameters. The engine never resolves to city or county level.

• Climate zone & moisture zone from climate_zones.json, per ASHRAE 169-2021 / IECC 2021. Eight zones (1=hot, 8=very cold) × three moisture regimes (A=humid, B=dry, C=marine).
• Degree days from degree_days.json, base 65°F, aggregated to a single representative HDD/CDD pair per climate zone. The lookup is climate-zone-resolved, not ZIP-resolved. Chelsea, MI uses CZ 5 values: HDD = 6,500 / CDD = 900.
• Electricity rate from electricity_rates.json, sourced from EIA Electric Power Monthly Table 5.6.A (residential, state-level), updated to February 2026. Users may override with their actual utility bill rate (engine accepts electricity_rate_override parameter in $/kWh).
• Gas rate from gas_rates.json, sourced from EIA Natural Gas Monthly state-level residential prices, updated February 2026, expressed in $/therm.
• Grid emission factor from egrid.json, EPA eGRID 2022 state-level STCO2RTA (output emission rate, CO₂-only, transmission losses excluded). Michigan = 0.458 kg CO₂/kWh.

Worked example (Chelsea, MI default): ZIP 48118 → state MI → CZ 5A → HDD 6,500 / CDD 900 / rate_elec = $0.200/kWh / rate_gas = $1.073/therm / EF_grid = 0.458 kg/kWh.

The building is modeled as a square-footprint rectangular volume. Window-to-wall ratio is fixed at 15% (close to RECS 2020 single-family national average of 14–18%). Wall height defaults to 9.0 ft. Door area defaults to a fixed 40 sf. Roof and floor areas equal the footprint (no slope or overhang adjustment).

Worked example: floor area 2,300 sf, 1 story → footprint 2,300 sf → side 48.0 ft → perimeter 191.8 ft → gross wall 1,726.6 sf → window 259 sf → net wall 1,427.5 sf → ceiling/floor 2,300 sf → volume 20,700 ft³. (Reproduces engine output exactly.)

Each element's heat-loss coefficient (UA) is area divided by effective R-value (or area times U-factor for fenestration). For walls, effective R is derived from the assembly: cavity R-value derated for thermal bridging through framing, plus continuous insulation R-value, plus a 1.0 air-film constant. For roofs, total assembly R-value plus the air-film constant. Window U-factor and door R are used directly.

Wall framing R-derate factors (fraction of cavity R that gets through the wall after thermal bridging):

Default R-values come from IECC 2021 prescriptive Table R402.1.2 for the climate zone unless an explicit assembly overrides them. CZ 5 code-min: R-20 cavity + R-5 continuous walls, R-60 ceiling, R-30 floor, U-0.30 windows, ACH50 = 3.0. Default door R = 5.0.

Slab-on-grade floor R is computed via a piecewise lookup, not directly from sub-slab insulation R-value. This accounts for ground thermal mass and soil temperature stabilization beneath the slab (most slab heat loss occurs through the perimeter, not the center). The lookup approximates the F-factor approach used by DOE Building America and ASHRAE 90.1 Appendix A:

Worked example (Chelsea code-min, 2×6 16" o.c. framing default): R_wall,eff = 20 × 0.70 + 5 + 1 = 20.0. R_ceiling,eff = 60 + 1 = 61.0. R-10 sub-slab → R_floor,eff = 45.0 (lookup). UA_wall = 1427.5 / 20 = 71.4. UA_ceiling = 2300 / 61 = 37.7. UA_floor = 2300 / 45 = 51.1. UA_window = 259 × 0.30 = 77.7. UA_door = 40 / 5 = 8.0. UA_inf = 20,700 × (3.0/20) × 0.018 = 55.9. UA_total = 301.8 BTU/hr·°F. (Reproduces engine output exactly.)

Annual heating energy uses the classic variable-base degree-day method (ASHRAE Handbook of Fundamentals Ch. 19), augmented with mechanical-ventilation losses and an internal-gains credit.

System efficiencies (constants, hardcoded):

HSPF is a seasonal metric expressed in BTU output per Wh input over the AHRI test season, so HSPF/3.412 yields the seasonal-average COP directly. This is the correct conversion for annual energy estimation (unlike SEER → COP, which is more nuanced — see Step 5).

Worked example (Chelsea, gas furnace): UA_total = 301.8, HDD = 6,500. Q_env = 301.8 × 6,500 × 24 = 47.08 MMBtu. Q_vent,cfm = 7.5 × 2.5 + 0.03 × 2,300 = 88 cfm. hours_heating = 6,500/20 × 24 = 7,800. q_vent = 88 × 60 × 30 × 0.018 × 1.0 = 2,851 BTU/hr. Q_vent = 22.18 MMBtu. gains_credit = (230 × 2.5 + 800) × 8,760 × (7,800/8,760) = 10.72 MMBtu. Q_net = 58.53 MMBtu. fuel_used = 58.53 MMBtu / 0.80 / 0.10 MMBtu/therm = 732 therms/yr. Compares to RECS 2020 Michigan median ~700 therms/yr for similar homes (within 5%).

Cooling load aggregates four sensible sources (envelope, solar gain through fenestration, internal gains, and ventilation), multiplies by a climate-zone-specific latent uplift factor for dehumidification, and converts to electricity using a seasonal COP.

Latent uplift LU_{zone,moisture} — dehumidification penalty applied to sensible cooling load. Lookup by (IECC climate zone) × (moisture zone):

SEER → COP conversion. SEER (Seasonal Energy Efficiency Ratio) is a part-load weighted seasonal metric (AHRI 210/240). The naive conversion COP = SEER/3.412 gives the peak-condition COP (the EER value at 95°F outdoor), which overstates real seasonal performance by ~12–15%. The engine uses EER ≈ 0.875 × SEER (DOE Building America practice) before converting to COP, giving SEER 14 → COP ≈ 3.59 rather than the naive 4.10.

Cooling system COPs (constants):

Worked example (Chelsea, code-min, SEER 14 central AC): UA_total = 301.8, CDD = 900. Q_env,c = 301.8 × 900 × 24 = 6.52 MMBtu. cool_days = 90. Q_solar = 259 × 0.40 × 200 × 4 × 90 = 7.46 MMBtu. hours_cool = 900/15 × 24 = 1,440. Q_int,c = 1,375 × 1,440 = 1.98 MMBtu. q_vent(cooling, ΔT=15) = 88 × 60 × 15 × 0.018 = 1,426 BTU/hr; Q_vent,c = 2.05 MMBtu. Q_sens = 18.0 MMBtu. LU_5A = 0.15 → Q_total,c = 20.7 MMBtu. cooling_kWh = 20,700,000 / (3.59 × 3,412) = 1,690 kWh/yr. (Reproduces engine output exactly.)

Energy factors (EF / UEF) for water heating systems, per DOE 10 CFR 430:

Worked example: 25 × 2.5 × 365 × 541 = 12.34 MMBtu demand. Gas tank: 12.34 / 0.62 / 0.10 = 199 therms/yr. HPWH: 12.34 / 3.5 / 3,412 = 1,034 kWh/yr. Baseline electricity: 5,500 × 1.0 × (1 + 0.0001 × 300) = 5,665 kWh/yr.

PV is treated as both an operational offset and an embodied carbon cost. The two halves use different accounting frameworks:

The max(0, …) clamp on net operational carbon is intentional. Per ISO 21930 attributional LCA, exported electricity does not create negative operational carbon for the building. Surplus generation displaces grid emissions somewhere, but that credit accrues to the grid operator or REC purchaser, not the building. This means oversizing PV beyond actual consumption adds embodied carbon (1.8 t per kW) without further offset — a 20 kW array on a leaky house has a higher whole-life total than a 10 kW array on the same house. The tool deliberately will not let users "buy negative carbon" with surplus PV.

Net metering simplification. The 70% retail-credit assumption for exported electricity is an approximate national average. Actual policies vary widely: full retail net metering (legacy CA NEM 1.0/2.0, MA, NJ) credits 100%; net-billing or avoided-cost programs (CA NEM 3.0, NV, MI DTE Distributed Generation) credit 25–50% of retail. The cost-savings number for PV-heavy scenarios will be optimistic in restrictive states (e.g., a Michigan DTE customer will see real savings of ~50–70% of the displayed value) and pessimistic in generous states. State-specific net-metering data is a planned improvement.

PV degradation. Real panels lose approximately 0.5%/year output through soiling, glass yellowing, and cell aging. Over 30 years this is ~14% cumulative degradation. The engine currently assumes constant generation, biasing PV scenarios optimistic by 5–7% over the full horizon.

Each envelope assembly (wall, roof, foundation) is built from a stack of materials drawn from the BEAM Tool v4.1 database (Builders for Climate Action). Each material has an EPD-derived A1–A3 emissions factor and, where applicable, a biogenic carbon storage value. Insulation materials are stored per square meter at RSI 1, so embodied scales linearly with thermal resistance.

Framing fractions (fraction of wall area occupied by solid lumber, including studs, headers, plates, corners): 2×4 16" o.c. = 0.25, 2×6 16" o.c. = 0.25, 2×6 24" o.c. = 0.22, advanced framing (OVE) 2×6 24" o.c. = 0.18, double-stud 2×4 24" o.c. = 0.30. These match published values from FHB, JLC, and OVE practice guides.

Windows use a flat 80 kg CO₂e/m² of glazed area, representative of mid-range residential aluminum-clad and vinyl windows. This is a known simplification: real values range from ~50 kg/m² (basic vinyl double-pane) to ~250 kg/m² (aluminum frame). Triple-pane wood-clad and double-pane vinyl currently report identical embodied carbon in this tool.

System boundary. Envelope embodied counts only EN 15978 stages A1 (raw materials), A2 (transport to factory), and A3 (manufacturing). Excluded: A4 site delivery, A5 installation, B1–B5 use and maintenance, B6 (operational energy, counted separately above), B7 (operational water, not relevant), C1–C4 disposal, D module credits. This is the conventional cradle-to-gate cutoff for residential WBLCA at this level of detail.

Interior partitions, mechanical/electrical/plumbing equipment, floor finishes, and cabinetry are estimated via per-square-foot category presets scaled by total floor area. This is the largest remaining source of absolute uncertainty in the whole-life total (±30% per component).

Finish-level presets (kg CO₂e per sf of floor area):

Sources: Carbon Leadership Forum MEP-LCA 2024 (MEP equipment); Athena Sustainable Materials Institute residential prototype outputs (partitions, finishes); Magwood (2022) Build Beyond Zero Ch. 8 (cabinetry, fixtures); BEAM Tool v4.1 (cross-check). These are typical-range averages, not project-specific. Real residential non-envelope carbon varies by ±30% from these presets depending on actual equipment specifications, finish brands, and interior configuration.

Whole-life total: embodied (envelope) + embodied (non-envelope) + embodied (PV) + 30-year operational. The 30-year horizon is conventional for residential WBLCA at this scope; it is not a building service-life claim. Real building service life is 60–100+ years.

Cost projection. The 4% central rate-increase assumption is consistent with historical nominal US residential energy inflation (~3–4%/yr over recent decades). The 0% / 6% bounds reflect realistic but uncertain trajectories. Note: these are nominal dollars, not real (inflation-adjusted). At ~2.5% general inflation, 4% nominal = ~1.5% real growth in energy expenditures. For consumer comparison ("how much will I pay over 30 years?"), nominal is the relevant metric. For NPV / discounted analysis, results would differ.

Two distinct uncertainty calculations run alongside every result. Absolute uncertainty reflects the realistic spread on a single scenario's total. Comparative uncertainty is much narrower because shared modeling assumptions cancel between scenarios — this is why "scenario A is 70% lower than scenario B" can be very-high-confidence even when each absolute value has a ±20% range.

Per-component uncertainty bands (one-sigma, expressed as fraction of central):

Statistical approach follows ISO 14044 §4.4.4 (uncertainty analysis), EPA WBLCA Guide (2019) §5.3 (comparative LCA noise floor), and standard practice for comparative residential LCA where shared boundary conditions reduce relative uncertainty. The 1.645σ multiplier corresponds to the 90% confidence interval assuming approximately normal error distribution on independent component uncertainties. The cost-uncertainty band (Step 10) is treated separately because rate-trajectory uncertainty is correlated across all scenarios (they share the same future utility-rate assumption); paired low-vs-low and high-vs-high comparisons are used for the savings range.

The list below is comprehensive and intentionally transparent. Items are flagged either as known omissions (a feature not yet implemented), simplifications (an approximation chosen for tractability), or directional biases (a known under- or over-estimate of a specific quantity). Items are listed in approximate order of magnitude of effect on the whole-life total.

1. Grid decarbonization trajectory not modeled. Operational carbon uses a static eGRID 2022 emission factor over the full 30-year horizon. Real grid carbon intensity is dropping at ~2%/year in most US regions and is projected to reach ~0.10 kg CO₂e/kWh nationally by 2055 (vs. ~0.43 today). Directional bias: all-electric and heat-pump scenarios are penalized; long-term electrification benefit is larger than shown. The tool is conservative for electrification.
2. Non-envelope embodied is preset-based, not project-specific. The largest single source of absolute uncertainty. Interior partitions, MEP equipment, floor finishes, and cabinetry use national-average per-square-foot factors at three finish levels (16 / 23 / 36 kg/sf). Real projects vary by ±30% depending on actual products. Type: simplification.
3. Heat pump COP does not degrade at low outdoor temperatures. Real cold-climate heat pumps lose 30–40% capacity and 20–30% COP at -5°F. The engine uses the constant seasonal HSPF-derived COP year-round. Directional bias: heat pump heating energy is underestimated in CZ 6–8 by approximately 15–25%. Most impactful for Anchorage, Duluth, and similar very-cold climates.

4. Replacement cycles (B-modules) not modeled. Vinyl cladding (~25-year service life) currently looks better on A1–A3 alone than fiber cement (~50+ years). Adding service-life modeling would reverse some current material rankings. Type: known omission. Affects cladding, roofing, and HVAC equipment comparisons.
5. PV degradation not modeled. Real panels lose ~0.5%/year output; over 30 years, ~14% cumulative. Engine assumes constant generation. Directional bias: PV-heavy scenarios optimistic by 5–7% on cumulative carbon offset and cost savings.
6. Climate-zone-aggregated HDD/CDD, not ZIP-level NOAA normals. Within CZ 5, Chelsea is ~6,300 HDD and Marquette is ~8,200 HDD, but both lookup to the same 6,500 HDD. Type: simplification. Effect: 1–3% spread on heating-dominated whole-life totals; largest for zones with wide internal variation.
7. Net metering 70% retail-credit is a national-average simplification. Real policies range from 100% (legacy CA, MA, NJ) to 25–50% (CA NEM 3.0, NV, MI DTE). Directional bias: cost savings for PV-heavy scenarios will be optimistic in restrictive states (a Michigan DTE customer should expect ~50–70% of displayed savings) and slightly pessimistic in generous states.
8. PV embodied does not vary by panel chemistry or module efficiency. All PV uses 1,800 kg CO₂e/kW (IEA-PVPS Task 12 median for c-Si). Real-world c-Si panels range 1,400–2,200 kg/kW; thin-film and emerging technologies differ further. Type: simplification.
9. Window embodied does not vary by frame or glazing. Flat 80 kg CO₂e/m² of glazed area regardless of vinyl/aluminum/wood/fiberglass frame or double/triple glazing. Real range is 50–250 kg/m². Type: simplification. Affects window upgrade comparisons.
10. Passive house preset assemblies do not scale with climate zone. The R-80 attic and R-36 walls used for the "Passive house" reference scenarios are calibrated for CZ 5–6. Generous for Miami (CZ 1), light for Anchorage (CZ 7–8). Directional bias: overestimates PH embodied in hot climates, underestimates PH operational savings in very-cold climates.
11. HRV/ERV efficiency is a single user-supplied number. No degradation, no temperature-dependent performance curve, no fan-power energy penalty. Type: simplification.

12. Hot water inlet temperature constant at 55°F. Real values range from ~40°F (Anchorage) to ~75°F (Florida). Type: simplification; partially cancels with HDD-based heating loads.
13. Heating-season ΔT fixed at 30°F. Used in mechanical ventilation heat-loss calculation. Real values vary from ~5°F (Miami) to ~50°F (Anchorage). Type: simplification; partially cancels with HDD-based envelope losses.
14. Geometry assumes square footprint, single shape. No accommodation for two-story, L-shaped, or complex envelopes. WWR fixed at 18% (RECS 2020 national average). Type: simplification.
15. Solar gain assumes average orientation. 200 BTU/sf·hr × 4 equivalent direct-exposure hours/day is an undifferentiated mix of N/S/E/W exposure. No glazing-orientation modeling. Type: simplification.
16. Infiltration N-factor fixed at 20. Real LBNL N-factors vary 15–25 depending on height, shielding, and climate. Type: simplification.
17. Dual biogenic reporting (gross + net) not yet implemented. Biogenic carbon storage in cellulose, wood fiber, and straw is currently shown as a single net credit. ISO 21930 requires reporting gross and net separately to handle different end-of-life assumptions. Type: known omission; affects how rigorously the tool can claim ISO 21930 compliance for biogenic materials.
18. Cost projection is nominal, not real (inflation-adjusted). 4%/yr central matches recent historical nominal energy inflation. Type: deliberate choice for consumer-facing comparison; sophisticated users running NPV analysis should apply their own discount rate.

These limitations are tracked openly because the tool's value comes from being honest about what it does and doesn't model. Most items above are planned improvements; some (e.g., constant grid carbon) are conservative simplifications that bias the tool against electrification, so fixing them strengthens rather than weakens the case for low-carbon design.

References

• ASHRAE 62.2-2022. Ventilation and Acceptable Indoor Air Quality in Residential Buildings. American Society of Heating, Refrigerating and Air-Conditioning Engineers.
• ASHRAE 169-2021. Climatic Data for Building Design Standards.
• ASHRAE Handbook of Fundamentals, 2021. Chapter 19 (Energy Estimating and Modeling) and Chapter 38 (Units and Conversions).
• ASHRAE 119. Air Leakage Performance for Detached Single-Family Residential Buildings.
• AHRI Standard 210/240-2023. Performance Rating of Unitary Air-Conditioning and Air-Source Heat Pump Equipment.
• Builders for Climate Action. BEAM Tool v4.1, 2024. Material database used for envelope embodied factors.
• Carbon Leadership Forum. MEP-LCA: Mechanical, Electrical, and Plumbing Life Cycle Assessment Report, 2024.
• DOE 10 CFR 430. Energy Conservation Program for Consumer Products: Test Procedures. Water heater Energy Factor / Uniform Energy Factor.
• EIA. Electric Power Monthly, Table 5.6.A residential rates by state. US Energy Information Administration; data version February 2026.
• EIA. Natural Gas Monthly, state-level residential prices. February 2026.
• EIA. Residential Energy Consumption Survey (RECS) 2020. Microdata used for baseline electricity, hot water, and heating fuel medians.
• EPA. Greenhouse Gas Emission Factors Hub, 2023 edition. CO₂, CH₄, N₂O factors for stationary residential combustion at AR5 GWP-100.
• EPA. eGRID 2022. US grid emission factors, state-level output emission rate (STCO2RTA).
• EPA. Building Life Cycle Assessment in Practice, 2019 (WBLCA Guide).
• IEA-PVPS Task 12. Life Cycle Inventories and Life Cycle Assessments of Photovoltaic Systems, 2020.
• International Code Council. International Energy Conservation Code (IECC) 2021, Table R402.1.2 (Insulation and Fenestration Requirements by Component).
• ISO 14044:2006. Environmental management — Life cycle assessment — Requirements and guidelines.
• ISO 21930:2017. Sustainability in buildings and civil engineering works — Core rules for environmental product declarations of construction products and services.
• IPCC AR5, 2014. Climate Change 2014: Synthesis Report. GWP-100 characterization factors.
• Magwood, C., 2022. Build Beyond Zero: New Ideas for Carbon-Smart Architecture. Island Press.
• NREL. PVWatts Calculator v8, 2024. State-level specific yield (kWh/kW·year).
• NREL. Cambium Standard Scenarios, 2023. Projected grid decarbonization by region.
• OMB Circular A-94. Guidelines and Discount Rates for Benefit-Cost Analysis of Federal Programs, Appendix C (revised). Real discount rate guidance.

The vertical bar at year 30 shows the 90% confidence interval on the absolute total. This is wider than the single number suggests because non-envelope materials, equipment, and finish choices have real product-level spread that this tool doesn't try to nail down precisely. A 200 t result might be 170–230 t in reality. That's not a flaw in the engine — it's an honest representation of what the data supports.

When comparing scenarios, the same modeling assumptions affect both — so the difference between them is much more reliable than the absolute values. "Very high confidence" means the gap is so large (3+ standard deviations) that the ranking is unambiguous. "Within modeling noise" means the difference falls within the inherent spread of the assumptions — the two scenarios are effectively a tie.