Uncertainty in predictions from air quality models

Abstract

This paper examines the role of the definition of the ensemble in determining the uncertainty in air quality model predictions. We difine the ensemble in terms of the variables included in the model input set. Because observations are also governed by variables not included in this set, we can have an infinitely large set of observations associated with a single model prediction. The variation of these observations about the model prediction corresponds to all possible values of the unknown variables, and it represents the uncertainty in the model prediction. This concept is illustrated by estimating the uncertainty related to model predictions for two dispersion problems.