Building type priors

The envelope and HVAC of your building structurally constrain what your indoor air can look like. The interpretation layer reads the building type at setup (detached, townhome, apartment, commercial) and applies priors that bias the AI toward the most likely explanations.

Detached single-family homes (most common in North American suburbs): largest envelope, most leakage (0.3 to 0.5 ACH passive), CO₂ rises slowest, ventilation often natural or via a single bath/kitchen fan. Outdoor pollutants infiltrate freely on windy days. Wood-burning is more common; gas appliances are usually outdoor-vented. The AI looks at high CO₂ with disbelief here (something is unusually sealed) and at sustained outdoor-tracking PM_2.5 with belief.

Townhomes and rowhouses: shared walls reduce envelope leakage, neighbor activity introduces cross-unit signals (smoke from a neighbor's gas range, dryer-vent backflow, fragrance from shared HVAC). Tight envelopes can produce surprisingly high CO₂ in bedrooms. Apartments: smallest air volume per person, highest occupancy density, often shared HVAC with neighbors (Building Air Loss patterns), high probability of cross-unit infiltration including cigarette smoke and cooking odors. LBNL infiltration measurements document the patterns.

Commercial buildings (offices, schools, retail): large HVAC systems with demand-controlled ventilation (see DCV), substantial outdoor-air dilution, but per-occupant rates vary widely. Cooking is concentrated in cafeterias rather than distributed. Cleaning happens on a fixed schedule (visible as nightly VOC pulses). The interpretation layer expects different signatures in each building type; a 1,500-ppm CO₂ reading in a detached home is unusual, in an apartment bedroom is normal, in a packed conference room is expected.