Towards improved models for indoor air chemistry: A Monte Carlo simulation study

Abstract

Model predictions are sensitive to a number of complex and often coupled input parameters. Some of these parameters have a wide range of acceptable values from literature and therefore choosing the appropriate value is non-trivial. In this paper, we use the INdoor Detailed Chemical Model (INDCM) to perform a Monte Carlo analysis, in which a wide but realistic range of model input parameter values is stochastically varied over 1000 model runs. The model output defines the likely range of the model performance, and directly correlates input parameter values with predicted indoor air species concentrations. The air exchange rate or the ozone deposition velocity onto internal materials such as painted walls, control the predicted concentrations of ozone, hydroxyl and peroxy radicals and peroxyacetyl nitrate species for our study conditions. The transmission of UV light from outdoors showed the strongest Spearman's rank positive correlation coefficients with predicted hydroxyl radical (0.92), and organic nitrate (0.95) concentrations. The deposition rate of ozone onto painted walls shows the strongest negative correlations with 4-oxopentanal (−0.86) and acetic acid (−0.83). Reducing the uncertainty around transmission of UV light indoors and ozone deposition rates onto surfaces reduces the model uncertainty by up to 70–80% for ozone and hydroxyl radical concentrations. Some species concentrations showed complex relationships with the various input parameters. For instance, maximum isoprene concentrations decreased with air exchange rate, but minimum isoprene concentrations were largely invariant. Emissions from human breath ensured that isoprene was always present in our model runs. However, its removal rate varied with the air exchange rate, which affected the concentrations of ozone and hydroxyl radicals (which can both chemically remove isoprene), and the direct removal rate by ventilation. Finally, we used our results to understand the 95% confidence bounds around our median predicted concentrations. For hydroxyl radicals, these were ±60% of the median value.