An assessment of residence scale residential PM2.5 exposure inaccuracy with a measurement-constrained theoretical model approach

Residences are an important micro-environment for human exposure to PM_2.5 mass. Existing measurement and simulation methodologies can return residence scale estimates of residential PM_2.5 mass concentration by averaging over the individual rooms of a residence. Such methods can be advantageous, however accuracy may be lost through averaging over rooms. Recent measurements allow the presented assessment of accuracy in exposure to residential PM_2.5 mass when using the residence scale. For multiple residences, measurements were taken in the kitchen, a child’s bedroom and either the living or dining room. Here we set a theoretical ground-truth for exposure of inhabitants to residential PM_2.5 mass by combining the measurements resolved across the three rooms with multiple plausible and informative occupancy schedules. Against this reference were compared three types of averaging over rooms for quantification of their accuracy: (i) equal-weighted and (ii) volume-weighted arithmetic means of multi-room measurements, and (iii) use of the living or dining room as representative of the whole residence. The error in residence scale estimation can lead to all residences found to have time-integrated PM_2.5 mass exposure above 100  through room-resolved estimation being miscategorised as having PM_2.5 less than 100 . Consequently, depending on the objective of work, room-averaging in deterministic studies may be deemed inappropriate. The median error in PM_2.5 daily exposure for typical room and residence occupancy schedules was 0 , and the median error for relative risk to all-cause mortality was 0, suggesting that room-averaging may be appropriate for population-level studies. A key implication for field studies and simulations is that researchers can combine this assessment with study objectives to identify whether the residence scale is sufficiently accurate or whether the room scale is required.