Recovery curves

After a source event ends, every pollutant decays toward its equilibrium concentration, but the shape of the decay differs by physics. The dashboard fits empirical curves to recent events in your room; the shape (not just the value) is what lets the AI distinguish a cooking event from a wildfire infiltration, even when the peak PM_2.5 is similar.

PM and CO₂ both decay roughly exponentially when the source stops and ventilation continues at a steady rate. Half-life is governed by total air changes per hour; see recovery time math. The PM curve is cleaner when ventilation dominates over settling (small particles, high airflow); coarse PM₁₀ often shows a faster initial drop as the heaviest fraction settles, then a slower exponential tail.

VOCs are bi-exponential. Indoor surfaces (walls, fabric, carpet) adsorb VOCs rapidly during a pulse, then desorb them slowly over hours to days. A 10-minute cooking event with 100 ppm-eq peak VOC index will drop most of the way back to baseline in 30 minutes, but the floor sits 10 to 20 above the original baseline for hours. The slow tail is the residual surface flux.

Humidity decays slowly compared to gases because water is also absorbed/desorbed by hygroscopic indoor materials (wood, paper, fabric). A shower humidity spike drops faster than it built (water is being removed by the bathroom fan), but the room takes longer to dry below its starting humidity than a dry-gas calculation would predict. Knowing these shapes by heart is how the AI says "that looks like a cooking event in recovery" with confidence rather than guessing.