Uncertainty in modelling PM10 and PM2.5 at a non-signalized traffic roundabout

Uncertainty in inputs to most air quality models of causal nature often results uncertainty in modelled concentrations as well. If incorporated, it may provide complete information on assured range of air pollutant levels. The study presents a sensitivity analysis of models and the probabilistic based estimates of uncertainties in their predictions. Two vehicular exhausts dispersion models have been used for forecasting hourly average PM10 and PM2.5 concentrations at a non–signalized roundabout traffic intersection. The uncertainties were estimated using first–and second–order Taylor series and Monte Carlo methods due to wind speed and wind direction and evaluated with one week particulate matter measurements (PM10 and PM2.5) during winter period. The amount of uncertainty due to wind speed was about 55% in the both models, resulted from wind direction was up to 5 to 20% of the modelled mean for receptors closer to the source but increased even up to 200% as the distance from the source increased. The uncertainty due to wind speed by second–order Taylor series matched with that by Monte Carlo method implying that a simple second–order Taylor series can be utilized for such studies instead of conventional time–taking Monte Carlo method.