DRAFT: NOT FOR CITATION OR DISTRIBUTION 1/23 Efficient Probabilistic Estimates of Surface Ozone Concentration Using an 1 Ensemble of Model Configurations and Direct Sensitivity Calculations

10 Because all models are a simplification of the phenomenon they aim to represent, it is 11 often more useful to estimate the probability of an event rather than a single “best” model 12 result. Previous air quality ensemble approaches have used computationally expensive 13 simulations of separately developed modeling systems. We present an efficient method 14 to generate ensembles with hundreds of members based on several structural 15 configurations of a single air quality modeling system. We use the Decoupled Direct 16 Method in 3D (DDM-3D) to directly calculate how ozone concentrations change as a 17 result of changes in input parameters. The modeled probability estimate is compared to 18 observations and is shown to have a high level of skill and improved resolution and 19 sharpness. This approach can help resolve the practical limits of incorporating 20 uncertainty estimation into deterministic air quality management modeling applications. 21