Monte Carlo estimates of uncertainties in outputs of regional ozone models due to uncertainties in inputs to the BEIS3 biogenics emissions model

A Monte Carlo (MC) probabilistic approach is used to estimate uncertainties in the outputs of three regional ozone models (MAQSIP, UAM-V, and URM) due to uncertainties in BEIS3 biogenics emissions model parameters and inputs. BEIS3 was developed by the EPA to estimate emissions of biogenic substances such as isoprene, monoterpenes, oxygenated and other volatile organic compounds (OVOCs), and biogenic nitric oxide (BNO). Uncertainties are investigated for three time periods - 24-29 May, 11-15 July and 4-9 September 1995, which have been extensively studied by others as part of emissions control planning exercises. In the MC approach, 1000 samples are drawn randomly and independently from seventeen BEIS3 parameters and inputs, whose distribution shapes and variances are determined from literature reviews. It is found that the 95 % confidence range on the calculated uncertainties in isoprene and BNO emissions cover approximately an order of magnitude, and that the 95 % confidence ranges on the calculated uncertainties in monoterpene and OVOC emissions cover about ± 20 %. The MC uncertainties in the predicted 1-hour and 8-hour averaged ozone concentrations were studied by drawing 20 random samples from the 1000 sets of BEIS3 outputs and running each regional ozone model 20 times for the three episodes. The estimated total uncertainties are found to be nearly the same for the three regional ozone models over the three time periods and for 1-hour and 8-hour averages. The 95 % confidence ranges on the uncertainties in calculated ozone concentrations, due only to uncertainties in BEIS3 inputs, cover approximately 15 to 20 % for all combinations of models, geographic domains, time periods, and averaging times.